Upload
ngotuyen
View
224
Download
0
Embed Size (px)
Citation preview
Hindawi Publishing CorporationJournal of Applied MathematicsVolume 2013 Article ID 960524 24 pageshttpdxdoiorg1011552013960524
Research ArticleModified Biogeography-Based Optimization withLocal Search Mechanism
Quanxi Feng12 Sanyang Liu1 Qunying Wu2 GuoQiang Tang2
Haomin Zhang2 and Huazhou Chen2
1 Department of Applied Mathematics Xidian University Xirsquoan 710071 China2 School of Science Guilin University of Technology Guilin 541004 China
Correspondence should be addressed to Quanxi Feng fqx9904gmailcom and Sanyang Liu liusanyan126com
Received 3 August 2013 Revised 6 September 2013 Accepted 10 September 2013
Academic Editor Ferenc Hartung
Copyright copy 2013 Quanxi Feng et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
Biogeography-based optimization (BBO) is a new effective population optimization algorithm based on the biogeography theorywith inherently insufficient exploration capability To address this limitation we proposed a modified BBO with local searchmechanism (denoted as MLBBO) In MLBBO a modified migration operator is integrated into BBO which can adopt moreinformation from other habitats to enhance the exploration ability Then a local search mechanism is used in BBO to supplementwithmodifiedmigration operator Extensive experimental tests are conducted on 27 benchmark functions to show the effectivenessof the proposed algorithm The simulation results have been compared with original BBO DE improved BBO algorithms andother evolutionary algorithms Finally the performance of the modified migration operator and local search mechanism are alsodiscussed
1 Introduction
In practical application many problems are regarded as opti-mization problems Several effective techniques have beendeveloped for solving optimization problems Traditionaltechniques [1 2] are effective methods for solving theseproblems but they need to know the property of problemssuch as continuity or differentiability In the past few decadesvarious evolutionary algorithms have been sprung up forsolving complex optimization problems for example geneticalgorithm (GA) [3] evolutionary programming (EP) [4]particle swarm optimization (PSO) [5] Ant Colony opti-mization (ACO) [6] differential evolution (DE) [7] andbiogeography-based optimization (BBO) [8] Comparedwithtraditional techniques evolutionary algorithms can solveoptimization problems without using some information suchas differentiability
Biogeography-based optimization (BBO) proposed bySimon [8] is a new entrant in the domain of globaloptimization based on the theory of biogeography BBO is
developed through simulating the emigration and immi-gration of species between habitats in the multidimen-sional solution space where each habitat represents a can-didate solution Just as species in biogeography migrateback and forth between habitats features in candidatesolutions are shared between solutions through migra-tion operator Good solutions tend to share their featureswith poor solutions However these features in the ori-gin good solution may exist in several solutions bothgood and poor solutions which may weaken explorationability
Several scholars have been working for enhancing theexploration ability To mention just a few examples Gonget al [9] presented a real-coded BBO (RCBBO) for contin-uous optimization problem through the use of real code torepresent the candidate solution and integration mutationoperator to improve the population diversity Cai et al [10]proposed hybrid BBO (BBO-EP) by combing evolutionaryprogramming with BBO Boussaıd et al [11] used BBO andDE to update the population alternately and gave a two-stage
2 Journal of Applied Mathematics
(1) Begin(2) For i = 1 to NP(3) Select119867
119894in the light of immigration rate 120582
119894
(4) If 119867119894is selected
(5) For j = 1 to NP(6) Select119867
119895according to emigration rate 120583
119895
(7) If rand (01) lt 120583119895
(8) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(9) EndIf(10) EndFor(11) EndIf(12) EndFor(13) End
Algorithm 1 Pseudocode of migration operator
algorithm (DBBO) Ergezer et al [12] employed opposition-based learning alongside BBO and developed oppositionalBBO (OBBO) Ma et al [13] incorporated resampling tech-nique into BBO for solving optimization problems undernoisy environments Lohokare et al [14] proposed a newvariant of BBO called aBBOmDE In the proposed algorithma modified mutation operator and clear duplicate operatorsare used to accelerate convergence speed and the modifiedDE (mDE) is embeded as a neighborhood search operator toimprove the fitness Gong et al [15] combined the explorationofDEwith the exploitation of BBO effectively and presented ahybrid DE with BBO namely DEBBO for global numericaloptimization problem Li et al [16] designed perturbingmigration operator based on the sinusoidal migration modeland advanced perturbing biogeography-based optimization(PBBO) Inspired by multiparent crossover in evolution-ary algorithm [17] Li and Yin [18] generalized migrationoperation based on multiparent crossover and presentedmultioperator biogeography-based optimization (MOBBO)These algorithms have acquired excellent performance
It is well known that both exploration ability and exploita-tion ability are two bases for population-based optimiza-tion algorithm However the exploration and exploitationare contradictory to each other [19] How to enhance theexploration ability is a key problem for improving theperformance of BBO For the sake of this inspired of DE [7]in this paper we modify the migration operator by applyinga modified mutation strategy to enhance the explorationability Meanwhile a local search mechanism is introducedin biogeography-based algorithm to supplement with themodifiedmigration operatorWename this BBOalgorithmasmodified biogeography-based optimization with local searchmechanism (denoted asMLBBO) Analysis and experimentalresults show that MLBBO with appropriate parameters canobtain excellent performance
The rest of this paper is organized as follows In Section 2we will review the basic BBO algorithm The modifiedBBO algorithm called MLBBO is presented and analyzedin Section 3 while experimental results are presented and
discussed in Section 4 Finally conclusions and future workare drawn in Section 5
2 Review of BBO
Biogeography-based optimization (BBO) [8] is a new emerg-ing population-based algorithm proposed through mimicspeciesmigration in natural biogeography InBBOalgorithmeach possible solution is a habitat and their features thatcharacterize suitability are called suitability index variables(SIVs) [10]The goodness of each solution is named as habitatsuitability index (HSI) In BBO a habitat H is a vector ofN values (SIVs) reaching global optima through migrationstep and mutation step In original migration information isshared among habitats based on the immigration rate 120582
119894and
emigration rate 120583119894which are linear functions of the number
of species in the habitat The linear model can be calculatedas follows
120582119894= 119868 (1 minus
119894
119899)
120583119894=119864119894
119899
(1)
where E is the maximum possible emigration rate I is themaximum possible immigration rate n is the maximumnumber of species and i is the number of species of the ithsolution
The pseudocode of migration operator [8] can be looselydescribed in Algorithm 1 Where NP (= n) is the populationsize
Because of ravenous predator or some other naturalcatastrophe [20] ecological equilibrium may be destroyedwhich could lead to drastically changing the species countand HSI of habitats This phenomenon can be modeled asSIV mutation and species count probabilities are used to
Journal of Applied Mathematics 3
(1) Begin(2) For i = 1 to NP(3) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(4) If 119867119894is selected
(5) For j = 1 to NP(6) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 NP(7) Select according to emigration rate 120583
119895
(8) If rand(01)lt 120583119895
(9) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(10) Else(11) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(12) EndIf(13) EndFor(14) EndIf(15) EndFor(16) End
Algorithm 2 Pseudocode of modified migration operator
determine mutation rates The mutation probability mi isexpressed as follows
119898119894= 119898max sdot (
1 minus 119901119894
119901max) (2)
where 119898max is a user-defined parameter and 119901max =
argmax 119901119894 119894 = 1 2 119899
In the originalmutation operator the SIV in each solutionis probabilistically replaced with a new feature randomlygenerated in the whole solution space which will tendto increase the diversity of the population Gong et al[9] modified this mutation operator and suggested threemutation operators Gaussian mutation operator Cauchymutation operator and Levy mutation operator used in realspace
3 Modified Biogeography-Based Optimizationwith Local Search Mechanism
31 ModifiedMigration Operator Migration operator is a keyoperator of original BBO which can improve the perfor-mance of BBO Meanwhile a habitat is modified throughmigration operator by simply replacing one of the SIV of thesimilar habitat whichmeans that the new habitat absorbs lessinformation from the others Therefore migration operatorlacks exploration ability
In order to absorb more information from other habitatsinspired of DE [7] the migration operator is modified byusing a DEmutation strategy DEbesttwomutation formulais used in this paper
DEbest2 119881119894= 119883best + 119865 sdot (1198831199031 minus 1198831199032) + 119865 sdot (1198831199033 minus 1198831199034)
(3)
where 119883best is the best solution F is scaling factor119881119894119894 isin 1 2 119873119875 is the ith trial vector and
1198831199031 1198831199032 1198831199033 and 119883
1199034are for mutually different solutions
BBO has acquired excellent performance since originalmigration operator can reserve good features in the popu-lation which are avoided being loosed or destroyed in theevolution process In order to keep the exploitation ability ofBBO we modified formula (3) as follows
BBObest2 119867119894 (SIV)
= 119867best (SIV) + 119865 sdot (1198671199031 (SIV) minus 1198671199032 (SIV))
+ 119865 sdot (1198671199033 (SIV) minus 1198671199034 (SIV))
(4)
where 119867best is the best habitat Hi is the immigration habi-tat and 119867
1199031 1198671199032 1198671199033 and 119867
1199034are four mutually different
habitats From this we can see from formula (4) thattwo differences are added to the new habitat and moreinformation from other habitats are adopted which canenhance the exploration ability Meanwhile it should benoted that parameter F plays an important role in balancingthe exploration ability and exploitation ability When F takes0 formula (4) absorbs information from best habitat itsexploitation ability enhanced When F increases formula(4) absorbs more information from other habitats andits exploration ability will increase but their exploitationability will decrease a certain degree Of course this sil-ver lining has its cloud Therefore F takes 05 in thispaper
In natural biogeography the migration model may benonlinear Hence Ma [21] introduced six migration modelsComparison results show that sinusoidal migration modelis the best one The sinusoidal migration model can becalculated as follows
120582119894=119868
2(1 + cos(119894120587
119899))
120583119894=119864
2(1 minus cos(119894120587
119899))
(5)
4 Journal of Applied Mathematics
(1) Begin(2) For i = 1 to NP(3) Compute the mutation probability p(4) Select variable119867
119894with probability p
(5) If 119867119894is selected
(6) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(7) Endif(8) Endfor(9) End
Algorithm 3 Pseudocode of Cauchy mutation operator
In order to enhance exploration ability we modifymigration operator described in Algorithm 1 by combiningformula 119867
119894(SIV) = 119867
119895(SIV) and formula (4) The modified
migration operator is described in Algorithm 2 We can see from Algorithm 2 that immigration habitats
not only accept information from best habitat but also fromBBObest2 On the one hand information fromHSI habitatsmake the operator maintain power exploitation ability Onthe other hand information from BBObest2 can enhancethe exploration ability From this we know that this modifiedmigration operator can preserve the exploitation ability andenhance the exploration ability simultaneously
32 Mutation with Mutation Operator Cataclysmic eventscan drastically change the HSI of a natural habitat andcause species count to differ from its equilibrium value Thisphenomenon can be modeled in BBO as SIV mutation andspecies count probabilities are used to determine mutationrates
In order to enhance the exploration ability of BBO amodified mutation operator Cauchy mutation operator isintegrated into BBO The probability density function ofCauchy distribution [9] is
119891119905 (119909) =
1
120587
119905
1199052 + 1199092 (6)
where 119909 isin 119877 and 119905 gt 0 is a scale parameterTheCauchymutation (t = 1) in this paper can be described
as
119867119894 (SIV) = 119867119894 (SIV) + 120575 (1) (7)
where 120575(1) is a randomness Cauchy value when parametert equals 1 In this paper we use the Cauchy distribution toupdate the solution which can be described in Algorithm 3
33 Local Search Mechanism As Simon has pointed out theconvergence speed is an obstacle for real application Sofinding mechanisms to speed up BBO could be an importantarea for further researchTherefore a local searchmechanismis proposed to accelerate the convergence speed of BBOIn order to remedy the insufficiency of modified migration
operator which mainly updates worse habitats selected byimmigration rate the local search mechanism is designedmainly for better habitats in the first half population
For each habitat119867119894in the first half population define
119867119894 (SIV) =
119867119894 (SIV) + 119886119897119901ℎ119886 lowast (119867119896 (SIV) minus 119867119894 (SIV))
119903119886119899119889 (0 1) lt 119901119897
119867119894 (SIV) else
(8)
where 119867119896denotes a habitat randomly selected from popula-
tion 119901119897denotes the local search probability and alpha is the
strength of local search
34 Selection Operator Selection operator is used to selectbetter habitats preserving in next generationwhich can impelthemetabolismof population and retain goodhabitats In thispaper greedy selection operator is used In other words theoriginal habitat119867
119894will be replaced by its corresponding trial
habitat119872119894if the HSI of the trial habitat119872
119894is better than that
of its original habitat119867119894
119867119894= 119867119894
if 119891 (119867119894) lt 119891 (119872
119894)
119872119894
else(9)
35 Main Procedure of MLBBO Exploration ability is mainlya contributory factor to step down the performance ofBBO algorithm In order to improve the performance ofBBO a modified BBO algorithm called MLBBO is pro-posed for global optimization problems in the continu-ous domain In MLBBO the original migration opera-tor and mutation operator are modified to enhance theexploration ability Moreover local search mechanism andselection operator are integrated into MLBBO to supple-ment with migration operator The pseudo-code of MLBBOis described in Algorithm 4 where NP is the populationsize
4 Experimental Study
41 Experimental Setting In this section several experimen-tal tests are carried out on 27 benchmark functions for
Journal of Applied Mathematics 5
(1) Begin(2) Initialize the population Pop with NP habitats randomly(3) Evaluate the fitness (HSI) for each Habitat119867
119894in Pop
(4) While the halting criteria are not satisfied do(5) Sort all the habitats from best to worst in line with their fitness values(6) Map the fitness to the number of species count S for each habitat(7) Calculate the immigration rate and emigration rate according to sinusoidal migration model(8) For i = 1 to NP lowastmigration stagelowast(9) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(10) If 119867119894is selected
(11) For j = 1 to NP(12) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 sdot sdot sdot 119873119875(13) Select119867
119895according to emigration rate 120583
119895
(14) If rand(01) lt 120583119895
(15) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(16) Else(17) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(18) EndIf(19) EndFor(20) EndIf(21) EndFor lowastend ofmigration stagelowast(22) For i = 1 to NP lowastmutation stagelowast(23) Compute the mutation probability 119901
119894
(24) Select variable119867119894with probability 119901
119894
(25) If 119867119894is selected
(26) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(27) Endif(28) EndFor lowastend ofmutation stagelowast(29) For i = 1 to NP2 lowastlocal search stagelowast(30) If rand(01) lt 119901
119897
(31) Select a habitat119867119896randomly
(32) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + alpha lowast (119867
119896(119878119868119881) minus 119867
119894(119878119868119881))
(33) EndIf(34) EndFor lowastend oflocal search stagelowast(35) Make sure each habitat legal based on boundary constraints(36) Use new generated habitat to replace duplicate(37) Evaluate the fitness (HSI) for trial habitat119872
119894in the new population Pop
(38) For i = 1 to NP lowastselection stagelowast(39) If119891(119872
119894) lt 119891(119867
119894)
(40) 119867119894= 119872119894
(41) EndIf(42) EndFor lowastend ofselection stagelowast(43) EndWhile(44) End
Algorithm 4 Pseudocode of MLBBO
validating the performance of the proposed algorithmThesefunctions consist of 13 standard benchmark functions [22](f1ndashf13) and 14 complex functions (F1ndashF14) from the set of testproblems listed in IEEE CEC 2005 [23] All the test functionsused in our experiments are high-dimensional functionswithchangeable dimension which are only briefly summarized inTable 1 A more detailed description of each function is givenin Table 8 (see also [22 23])
For fair comparison we have chosen a reasonable set ofparameter values For all experimental simulations in this
paper we will use the following parameters setting unless achange is mentioned
(1) population size NP = 100(2) maximum immigration rate I =1(3) maximum emigration rate E =1(4) mutation probability119898max = 0001(5) value to reach (VTR) = 10minus6 except for f7 F6ndashF14 of
VTR = 10minus2
6 Journal of Applied Mathematics
Table 1 Test unctions used in our experimental tests
Function name Dimension Search Space Optimaf1 Sphere n [minus100 100] 0f2 Schwefelrsquos Problem 222 n [10 10] 0f3 Schwefelrsquos Problem 12 n [minus100 100] 0f4 Schwefelrsquos Problem 221 n [minus100 100] 0f5 Rosenbrock function n [minus30 30] 0f6 Step function n [minus100 100] 0f7 Quartic function n [minus128 128] 0f8 Schwefel function n [minus512 512] minus125695f9 Rastrigin function n [minus512 512] 0f10 Ackley function n [minus32 32] 0f11 Griewank function n [minus600 600] 0f12 Penalized function 1 n [minus50 50] 0f13 Penalized function 2 n [minus50 50] 0F1 Shifted sphere function n [minus100 100] minus450F2 Shifted Schwefelrsquos Problem 12 n [minus100 100] minus450F3 Shifted rotated high conditioned elliptic function n [minus100 100] minus450F4 Shifted Schwefelrsquos Problem 12 with noise in fitness n [minus100 100] minus450F5 Schwefelrsquos Problem 26 with global optimum on bounds n [minus30 30] minus310F6 Shifted Rosenbrockrsquos function n [minus100 100] 390F7 Shifted rotated Griewankrsquos function without bounds n Initialize in [0 600] minus180F8 Shifted rotated Ackleyrsquos function with global optimum on bounds n [minus32 32] minus140F9 Shifted Rastriginrsquos function n [minus512 512] minus330F10 Shifted Rotated Rastriginrsquos function n [minus512 512] minus330F11 Shifted Rotated Weierstrass function n [minus600 600] 90F12 Schwefelrsquos Problem 213 n [minus100 100] minus460F13 Expanded extended Griewankrsquos plus Rosenbrockrsquos function (F8F2) n [minus50 50] minus130F14 Shifted rotated expanded scafferrsquos F6 n [minus100 100] minus300
(6) maximum number of fitness function evaluations(MaxNFFEs) 150000 for f1 f6 f10 f12 and f13200000 for f2 and f11 300000 for f7 f8 f9 F1ndashF14and 500000 for f3 f4 and f5
In our experimental tests all the functions are optimizedover 30 independent runs All experimental tests are carriedout on a computer with 186GHz Intel core Processor 1 GB ofRAM and Window XP operation system in Matlab software76
42 Effect of the Strength Alpha and Rate 119901119897on the Perfor-
mance of MLBBO In this section we investigate the impactof local search strength alpha and search probability 119901
119897on
the proposed algorithm In order to analyze the performancemore scientific eight functions are chosen from Table 1Theyare Sphere Schwefel 12 Schwefel Penalized2 F1 F3 F10 and
F13 which belong to standard unimodal function standardmultimodal function shifted function shifted rotated func-tion and expanded function respectively Search strengthalpha are 03 05 08 and 11 Search rates 119901
119897are 0 02 04
06 08 and 1 It is important to note that 119901119897equal to 0
means that local search mechanism is not conducted Hencein our experiments alpha and 119901
119897are tested according to
24 situations MLBBO algorithm executes 30 independentruns on each function and convergence figures of averagefitness values (which are function error values) are plotted inFigure 1
From Figure 1 we can observe that different search rate 119901119897
have different results When 119901119897is 0 the results are worse than
the others which means that local search mechanism canimprove the performance of proposed algorithm When 119901
119897is
04 we obtain a faster convergence speed and better resultson the Sphere function For the other seven test functions
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
2 Journal of Applied Mathematics
(1) Begin(2) For i = 1 to NP(3) Select119867
119894in the light of immigration rate 120582
119894
(4) If 119867119894is selected
(5) For j = 1 to NP(6) Select119867
119895according to emigration rate 120583
119895
(7) If rand (01) lt 120583119895
(8) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(9) EndIf(10) EndFor(11) EndIf(12) EndFor(13) End
Algorithm 1 Pseudocode of migration operator
algorithm (DBBO) Ergezer et al [12] employed opposition-based learning alongside BBO and developed oppositionalBBO (OBBO) Ma et al [13] incorporated resampling tech-nique into BBO for solving optimization problems undernoisy environments Lohokare et al [14] proposed a newvariant of BBO called aBBOmDE In the proposed algorithma modified mutation operator and clear duplicate operatorsare used to accelerate convergence speed and the modifiedDE (mDE) is embeded as a neighborhood search operator toimprove the fitness Gong et al [15] combined the explorationofDEwith the exploitation of BBO effectively and presented ahybrid DE with BBO namely DEBBO for global numericaloptimization problem Li et al [16] designed perturbingmigration operator based on the sinusoidal migration modeland advanced perturbing biogeography-based optimization(PBBO) Inspired by multiparent crossover in evolution-ary algorithm [17] Li and Yin [18] generalized migrationoperation based on multiparent crossover and presentedmultioperator biogeography-based optimization (MOBBO)These algorithms have acquired excellent performance
It is well known that both exploration ability and exploita-tion ability are two bases for population-based optimiza-tion algorithm However the exploration and exploitationare contradictory to each other [19] How to enhance theexploration ability is a key problem for improving theperformance of BBO For the sake of this inspired of DE [7]in this paper we modify the migration operator by applyinga modified mutation strategy to enhance the explorationability Meanwhile a local search mechanism is introducedin biogeography-based algorithm to supplement with themodifiedmigration operatorWename this BBOalgorithmasmodified biogeography-based optimization with local searchmechanism (denoted asMLBBO) Analysis and experimentalresults show that MLBBO with appropriate parameters canobtain excellent performance
The rest of this paper is organized as follows In Section 2we will review the basic BBO algorithm The modifiedBBO algorithm called MLBBO is presented and analyzedin Section 3 while experimental results are presented and
discussed in Section 4 Finally conclusions and future workare drawn in Section 5
2 Review of BBO
Biogeography-based optimization (BBO) [8] is a new emerg-ing population-based algorithm proposed through mimicspeciesmigration in natural biogeography InBBOalgorithmeach possible solution is a habitat and their features thatcharacterize suitability are called suitability index variables(SIVs) [10]The goodness of each solution is named as habitatsuitability index (HSI) In BBO a habitat H is a vector ofN values (SIVs) reaching global optima through migrationstep and mutation step In original migration information isshared among habitats based on the immigration rate 120582
119894and
emigration rate 120583119894which are linear functions of the number
of species in the habitat The linear model can be calculatedas follows
120582119894= 119868 (1 minus
119894
119899)
120583119894=119864119894
119899
(1)
where E is the maximum possible emigration rate I is themaximum possible immigration rate n is the maximumnumber of species and i is the number of species of the ithsolution
The pseudocode of migration operator [8] can be looselydescribed in Algorithm 1 Where NP (= n) is the populationsize
Because of ravenous predator or some other naturalcatastrophe [20] ecological equilibrium may be destroyedwhich could lead to drastically changing the species countand HSI of habitats This phenomenon can be modeled asSIV mutation and species count probabilities are used to
Journal of Applied Mathematics 3
(1) Begin(2) For i = 1 to NP(3) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(4) If 119867119894is selected
(5) For j = 1 to NP(6) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 NP(7) Select according to emigration rate 120583
119895
(8) If rand(01)lt 120583119895
(9) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(10) Else(11) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(12) EndIf(13) EndFor(14) EndIf(15) EndFor(16) End
Algorithm 2 Pseudocode of modified migration operator
determine mutation rates The mutation probability mi isexpressed as follows
119898119894= 119898max sdot (
1 minus 119901119894
119901max) (2)
where 119898max is a user-defined parameter and 119901max =
argmax 119901119894 119894 = 1 2 119899
In the originalmutation operator the SIV in each solutionis probabilistically replaced with a new feature randomlygenerated in the whole solution space which will tendto increase the diversity of the population Gong et al[9] modified this mutation operator and suggested threemutation operators Gaussian mutation operator Cauchymutation operator and Levy mutation operator used in realspace
3 Modified Biogeography-Based Optimizationwith Local Search Mechanism
31 ModifiedMigration Operator Migration operator is a keyoperator of original BBO which can improve the perfor-mance of BBO Meanwhile a habitat is modified throughmigration operator by simply replacing one of the SIV of thesimilar habitat whichmeans that the new habitat absorbs lessinformation from the others Therefore migration operatorlacks exploration ability
In order to absorb more information from other habitatsinspired of DE [7] the migration operator is modified byusing a DEmutation strategy DEbesttwomutation formulais used in this paper
DEbest2 119881119894= 119883best + 119865 sdot (1198831199031 minus 1198831199032) + 119865 sdot (1198831199033 minus 1198831199034)
(3)
where 119883best is the best solution F is scaling factor119881119894119894 isin 1 2 119873119875 is the ith trial vector and
1198831199031 1198831199032 1198831199033 and 119883
1199034are for mutually different solutions
BBO has acquired excellent performance since originalmigration operator can reserve good features in the popu-lation which are avoided being loosed or destroyed in theevolution process In order to keep the exploitation ability ofBBO we modified formula (3) as follows
BBObest2 119867119894 (SIV)
= 119867best (SIV) + 119865 sdot (1198671199031 (SIV) minus 1198671199032 (SIV))
+ 119865 sdot (1198671199033 (SIV) minus 1198671199034 (SIV))
(4)
where 119867best is the best habitat Hi is the immigration habi-tat and 119867
1199031 1198671199032 1198671199033 and 119867
1199034are four mutually different
habitats From this we can see from formula (4) thattwo differences are added to the new habitat and moreinformation from other habitats are adopted which canenhance the exploration ability Meanwhile it should benoted that parameter F plays an important role in balancingthe exploration ability and exploitation ability When F takes0 formula (4) absorbs information from best habitat itsexploitation ability enhanced When F increases formula(4) absorbs more information from other habitats andits exploration ability will increase but their exploitationability will decrease a certain degree Of course this sil-ver lining has its cloud Therefore F takes 05 in thispaper
In natural biogeography the migration model may benonlinear Hence Ma [21] introduced six migration modelsComparison results show that sinusoidal migration modelis the best one The sinusoidal migration model can becalculated as follows
120582119894=119868
2(1 + cos(119894120587
119899))
120583119894=119864
2(1 minus cos(119894120587
119899))
(5)
4 Journal of Applied Mathematics
(1) Begin(2) For i = 1 to NP(3) Compute the mutation probability p(4) Select variable119867
119894with probability p
(5) If 119867119894is selected
(6) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(7) Endif(8) Endfor(9) End
Algorithm 3 Pseudocode of Cauchy mutation operator
In order to enhance exploration ability we modifymigration operator described in Algorithm 1 by combiningformula 119867
119894(SIV) = 119867
119895(SIV) and formula (4) The modified
migration operator is described in Algorithm 2 We can see from Algorithm 2 that immigration habitats
not only accept information from best habitat but also fromBBObest2 On the one hand information fromHSI habitatsmake the operator maintain power exploitation ability Onthe other hand information from BBObest2 can enhancethe exploration ability From this we know that this modifiedmigration operator can preserve the exploitation ability andenhance the exploration ability simultaneously
32 Mutation with Mutation Operator Cataclysmic eventscan drastically change the HSI of a natural habitat andcause species count to differ from its equilibrium value Thisphenomenon can be modeled in BBO as SIV mutation andspecies count probabilities are used to determine mutationrates
In order to enhance the exploration ability of BBO amodified mutation operator Cauchy mutation operator isintegrated into BBO The probability density function ofCauchy distribution [9] is
119891119905 (119909) =
1
120587
119905
1199052 + 1199092 (6)
where 119909 isin 119877 and 119905 gt 0 is a scale parameterTheCauchymutation (t = 1) in this paper can be described
as
119867119894 (SIV) = 119867119894 (SIV) + 120575 (1) (7)
where 120575(1) is a randomness Cauchy value when parametert equals 1 In this paper we use the Cauchy distribution toupdate the solution which can be described in Algorithm 3
33 Local Search Mechanism As Simon has pointed out theconvergence speed is an obstacle for real application Sofinding mechanisms to speed up BBO could be an importantarea for further researchTherefore a local searchmechanismis proposed to accelerate the convergence speed of BBOIn order to remedy the insufficiency of modified migration
operator which mainly updates worse habitats selected byimmigration rate the local search mechanism is designedmainly for better habitats in the first half population
For each habitat119867119894in the first half population define
119867119894 (SIV) =
119867119894 (SIV) + 119886119897119901ℎ119886 lowast (119867119896 (SIV) minus 119867119894 (SIV))
119903119886119899119889 (0 1) lt 119901119897
119867119894 (SIV) else
(8)
where 119867119896denotes a habitat randomly selected from popula-
tion 119901119897denotes the local search probability and alpha is the
strength of local search
34 Selection Operator Selection operator is used to selectbetter habitats preserving in next generationwhich can impelthemetabolismof population and retain goodhabitats In thispaper greedy selection operator is used In other words theoriginal habitat119867
119894will be replaced by its corresponding trial
habitat119872119894if the HSI of the trial habitat119872
119894is better than that
of its original habitat119867119894
119867119894= 119867119894
if 119891 (119867119894) lt 119891 (119872
119894)
119872119894
else(9)
35 Main Procedure of MLBBO Exploration ability is mainlya contributory factor to step down the performance ofBBO algorithm In order to improve the performance ofBBO a modified BBO algorithm called MLBBO is pro-posed for global optimization problems in the continu-ous domain In MLBBO the original migration opera-tor and mutation operator are modified to enhance theexploration ability Moreover local search mechanism andselection operator are integrated into MLBBO to supple-ment with migration operator The pseudo-code of MLBBOis described in Algorithm 4 where NP is the populationsize
4 Experimental Study
41 Experimental Setting In this section several experimen-tal tests are carried out on 27 benchmark functions for
Journal of Applied Mathematics 5
(1) Begin(2) Initialize the population Pop with NP habitats randomly(3) Evaluate the fitness (HSI) for each Habitat119867
119894in Pop
(4) While the halting criteria are not satisfied do(5) Sort all the habitats from best to worst in line with their fitness values(6) Map the fitness to the number of species count S for each habitat(7) Calculate the immigration rate and emigration rate according to sinusoidal migration model(8) For i = 1 to NP lowastmigration stagelowast(9) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(10) If 119867119894is selected
(11) For j = 1 to NP(12) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 sdot sdot sdot 119873119875(13) Select119867
119895according to emigration rate 120583
119895
(14) If rand(01) lt 120583119895
(15) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(16) Else(17) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(18) EndIf(19) EndFor(20) EndIf(21) EndFor lowastend ofmigration stagelowast(22) For i = 1 to NP lowastmutation stagelowast(23) Compute the mutation probability 119901
119894
(24) Select variable119867119894with probability 119901
119894
(25) If 119867119894is selected
(26) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(27) Endif(28) EndFor lowastend ofmutation stagelowast(29) For i = 1 to NP2 lowastlocal search stagelowast(30) If rand(01) lt 119901
119897
(31) Select a habitat119867119896randomly
(32) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + alpha lowast (119867
119896(119878119868119881) minus 119867
119894(119878119868119881))
(33) EndIf(34) EndFor lowastend oflocal search stagelowast(35) Make sure each habitat legal based on boundary constraints(36) Use new generated habitat to replace duplicate(37) Evaluate the fitness (HSI) for trial habitat119872
119894in the new population Pop
(38) For i = 1 to NP lowastselection stagelowast(39) If119891(119872
119894) lt 119891(119867
119894)
(40) 119867119894= 119872119894
(41) EndIf(42) EndFor lowastend ofselection stagelowast(43) EndWhile(44) End
Algorithm 4 Pseudocode of MLBBO
validating the performance of the proposed algorithmThesefunctions consist of 13 standard benchmark functions [22](f1ndashf13) and 14 complex functions (F1ndashF14) from the set of testproblems listed in IEEE CEC 2005 [23] All the test functionsused in our experiments are high-dimensional functionswithchangeable dimension which are only briefly summarized inTable 1 A more detailed description of each function is givenin Table 8 (see also [22 23])
For fair comparison we have chosen a reasonable set ofparameter values For all experimental simulations in this
paper we will use the following parameters setting unless achange is mentioned
(1) population size NP = 100(2) maximum immigration rate I =1(3) maximum emigration rate E =1(4) mutation probability119898max = 0001(5) value to reach (VTR) = 10minus6 except for f7 F6ndashF14 of
VTR = 10minus2
6 Journal of Applied Mathematics
Table 1 Test unctions used in our experimental tests
Function name Dimension Search Space Optimaf1 Sphere n [minus100 100] 0f2 Schwefelrsquos Problem 222 n [10 10] 0f3 Schwefelrsquos Problem 12 n [minus100 100] 0f4 Schwefelrsquos Problem 221 n [minus100 100] 0f5 Rosenbrock function n [minus30 30] 0f6 Step function n [minus100 100] 0f7 Quartic function n [minus128 128] 0f8 Schwefel function n [minus512 512] minus125695f9 Rastrigin function n [minus512 512] 0f10 Ackley function n [minus32 32] 0f11 Griewank function n [minus600 600] 0f12 Penalized function 1 n [minus50 50] 0f13 Penalized function 2 n [minus50 50] 0F1 Shifted sphere function n [minus100 100] minus450F2 Shifted Schwefelrsquos Problem 12 n [minus100 100] minus450F3 Shifted rotated high conditioned elliptic function n [minus100 100] minus450F4 Shifted Schwefelrsquos Problem 12 with noise in fitness n [minus100 100] minus450F5 Schwefelrsquos Problem 26 with global optimum on bounds n [minus30 30] minus310F6 Shifted Rosenbrockrsquos function n [minus100 100] 390F7 Shifted rotated Griewankrsquos function without bounds n Initialize in [0 600] minus180F8 Shifted rotated Ackleyrsquos function with global optimum on bounds n [minus32 32] minus140F9 Shifted Rastriginrsquos function n [minus512 512] minus330F10 Shifted Rotated Rastriginrsquos function n [minus512 512] minus330F11 Shifted Rotated Weierstrass function n [minus600 600] 90F12 Schwefelrsquos Problem 213 n [minus100 100] minus460F13 Expanded extended Griewankrsquos plus Rosenbrockrsquos function (F8F2) n [minus50 50] minus130F14 Shifted rotated expanded scafferrsquos F6 n [minus100 100] minus300
(6) maximum number of fitness function evaluations(MaxNFFEs) 150000 for f1 f6 f10 f12 and f13200000 for f2 and f11 300000 for f7 f8 f9 F1ndashF14and 500000 for f3 f4 and f5
In our experimental tests all the functions are optimizedover 30 independent runs All experimental tests are carriedout on a computer with 186GHz Intel core Processor 1 GB ofRAM and Window XP operation system in Matlab software76
42 Effect of the Strength Alpha and Rate 119901119897on the Perfor-
mance of MLBBO In this section we investigate the impactof local search strength alpha and search probability 119901
119897on
the proposed algorithm In order to analyze the performancemore scientific eight functions are chosen from Table 1Theyare Sphere Schwefel 12 Schwefel Penalized2 F1 F3 F10 and
F13 which belong to standard unimodal function standardmultimodal function shifted function shifted rotated func-tion and expanded function respectively Search strengthalpha are 03 05 08 and 11 Search rates 119901
119897are 0 02 04
06 08 and 1 It is important to note that 119901119897equal to 0
means that local search mechanism is not conducted Hencein our experiments alpha and 119901
119897are tested according to
24 situations MLBBO algorithm executes 30 independentruns on each function and convergence figures of averagefitness values (which are function error values) are plotted inFigure 1
From Figure 1 we can observe that different search rate 119901119897
have different results When 119901119897is 0 the results are worse than
the others which means that local search mechanism canimprove the performance of proposed algorithm When 119901
119897is
04 we obtain a faster convergence speed and better resultson the Sphere function For the other seven test functions
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 3
(1) Begin(2) For i = 1 to NP(3) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(4) If 119867119894is selected
(5) For j = 1 to NP(6) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 NP(7) Select according to emigration rate 120583
119895
(8) If rand(01)lt 120583119895
(9) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(10) Else(11) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(12) EndIf(13) EndFor(14) EndIf(15) EndFor(16) End
Algorithm 2 Pseudocode of modified migration operator
determine mutation rates The mutation probability mi isexpressed as follows
119898119894= 119898max sdot (
1 minus 119901119894
119901max) (2)
where 119898max is a user-defined parameter and 119901max =
argmax 119901119894 119894 = 1 2 119899
In the originalmutation operator the SIV in each solutionis probabilistically replaced with a new feature randomlygenerated in the whole solution space which will tendto increase the diversity of the population Gong et al[9] modified this mutation operator and suggested threemutation operators Gaussian mutation operator Cauchymutation operator and Levy mutation operator used in realspace
3 Modified Biogeography-Based Optimizationwith Local Search Mechanism
31 ModifiedMigration Operator Migration operator is a keyoperator of original BBO which can improve the perfor-mance of BBO Meanwhile a habitat is modified throughmigration operator by simply replacing one of the SIV of thesimilar habitat whichmeans that the new habitat absorbs lessinformation from the others Therefore migration operatorlacks exploration ability
In order to absorb more information from other habitatsinspired of DE [7] the migration operator is modified byusing a DEmutation strategy DEbesttwomutation formulais used in this paper
DEbest2 119881119894= 119883best + 119865 sdot (1198831199031 minus 1198831199032) + 119865 sdot (1198831199033 minus 1198831199034)
(3)
where 119883best is the best solution F is scaling factor119881119894119894 isin 1 2 119873119875 is the ith trial vector and
1198831199031 1198831199032 1198831199033 and 119883
1199034are for mutually different solutions
BBO has acquired excellent performance since originalmigration operator can reserve good features in the popu-lation which are avoided being loosed or destroyed in theevolution process In order to keep the exploitation ability ofBBO we modified formula (3) as follows
BBObest2 119867119894 (SIV)
= 119867best (SIV) + 119865 sdot (1198671199031 (SIV) minus 1198671199032 (SIV))
+ 119865 sdot (1198671199033 (SIV) minus 1198671199034 (SIV))
(4)
where 119867best is the best habitat Hi is the immigration habi-tat and 119867
1199031 1198671199032 1198671199033 and 119867
1199034are four mutually different
habitats From this we can see from formula (4) thattwo differences are added to the new habitat and moreinformation from other habitats are adopted which canenhance the exploration ability Meanwhile it should benoted that parameter F plays an important role in balancingthe exploration ability and exploitation ability When F takes0 formula (4) absorbs information from best habitat itsexploitation ability enhanced When F increases formula(4) absorbs more information from other habitats andits exploration ability will increase but their exploitationability will decrease a certain degree Of course this sil-ver lining has its cloud Therefore F takes 05 in thispaper
In natural biogeography the migration model may benonlinear Hence Ma [21] introduced six migration modelsComparison results show that sinusoidal migration modelis the best one The sinusoidal migration model can becalculated as follows
120582119894=119868
2(1 + cos(119894120587
119899))
120583119894=119864
2(1 minus cos(119894120587
119899))
(5)
4 Journal of Applied Mathematics
(1) Begin(2) For i = 1 to NP(3) Compute the mutation probability p(4) Select variable119867
119894with probability p
(5) If 119867119894is selected
(6) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(7) Endif(8) Endfor(9) End
Algorithm 3 Pseudocode of Cauchy mutation operator
In order to enhance exploration ability we modifymigration operator described in Algorithm 1 by combiningformula 119867
119894(SIV) = 119867
119895(SIV) and formula (4) The modified
migration operator is described in Algorithm 2 We can see from Algorithm 2 that immigration habitats
not only accept information from best habitat but also fromBBObest2 On the one hand information fromHSI habitatsmake the operator maintain power exploitation ability Onthe other hand information from BBObest2 can enhancethe exploration ability From this we know that this modifiedmigration operator can preserve the exploitation ability andenhance the exploration ability simultaneously
32 Mutation with Mutation Operator Cataclysmic eventscan drastically change the HSI of a natural habitat andcause species count to differ from its equilibrium value Thisphenomenon can be modeled in BBO as SIV mutation andspecies count probabilities are used to determine mutationrates
In order to enhance the exploration ability of BBO amodified mutation operator Cauchy mutation operator isintegrated into BBO The probability density function ofCauchy distribution [9] is
119891119905 (119909) =
1
120587
119905
1199052 + 1199092 (6)
where 119909 isin 119877 and 119905 gt 0 is a scale parameterTheCauchymutation (t = 1) in this paper can be described
as
119867119894 (SIV) = 119867119894 (SIV) + 120575 (1) (7)
where 120575(1) is a randomness Cauchy value when parametert equals 1 In this paper we use the Cauchy distribution toupdate the solution which can be described in Algorithm 3
33 Local Search Mechanism As Simon has pointed out theconvergence speed is an obstacle for real application Sofinding mechanisms to speed up BBO could be an importantarea for further researchTherefore a local searchmechanismis proposed to accelerate the convergence speed of BBOIn order to remedy the insufficiency of modified migration
operator which mainly updates worse habitats selected byimmigration rate the local search mechanism is designedmainly for better habitats in the first half population
For each habitat119867119894in the first half population define
119867119894 (SIV) =
119867119894 (SIV) + 119886119897119901ℎ119886 lowast (119867119896 (SIV) minus 119867119894 (SIV))
119903119886119899119889 (0 1) lt 119901119897
119867119894 (SIV) else
(8)
where 119867119896denotes a habitat randomly selected from popula-
tion 119901119897denotes the local search probability and alpha is the
strength of local search
34 Selection Operator Selection operator is used to selectbetter habitats preserving in next generationwhich can impelthemetabolismof population and retain goodhabitats In thispaper greedy selection operator is used In other words theoriginal habitat119867
119894will be replaced by its corresponding trial
habitat119872119894if the HSI of the trial habitat119872
119894is better than that
of its original habitat119867119894
119867119894= 119867119894
if 119891 (119867119894) lt 119891 (119872
119894)
119872119894
else(9)
35 Main Procedure of MLBBO Exploration ability is mainlya contributory factor to step down the performance ofBBO algorithm In order to improve the performance ofBBO a modified BBO algorithm called MLBBO is pro-posed for global optimization problems in the continu-ous domain In MLBBO the original migration opera-tor and mutation operator are modified to enhance theexploration ability Moreover local search mechanism andselection operator are integrated into MLBBO to supple-ment with migration operator The pseudo-code of MLBBOis described in Algorithm 4 where NP is the populationsize
4 Experimental Study
41 Experimental Setting In this section several experimen-tal tests are carried out on 27 benchmark functions for
Journal of Applied Mathematics 5
(1) Begin(2) Initialize the population Pop with NP habitats randomly(3) Evaluate the fitness (HSI) for each Habitat119867
119894in Pop
(4) While the halting criteria are not satisfied do(5) Sort all the habitats from best to worst in line with their fitness values(6) Map the fitness to the number of species count S for each habitat(7) Calculate the immigration rate and emigration rate according to sinusoidal migration model(8) For i = 1 to NP lowastmigration stagelowast(9) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(10) If 119867119894is selected
(11) For j = 1 to NP(12) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 sdot sdot sdot 119873119875(13) Select119867
119895according to emigration rate 120583
119895
(14) If rand(01) lt 120583119895
(15) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(16) Else(17) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(18) EndIf(19) EndFor(20) EndIf(21) EndFor lowastend ofmigration stagelowast(22) For i = 1 to NP lowastmutation stagelowast(23) Compute the mutation probability 119901
119894
(24) Select variable119867119894with probability 119901
119894
(25) If 119867119894is selected
(26) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(27) Endif(28) EndFor lowastend ofmutation stagelowast(29) For i = 1 to NP2 lowastlocal search stagelowast(30) If rand(01) lt 119901
119897
(31) Select a habitat119867119896randomly
(32) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + alpha lowast (119867
119896(119878119868119881) minus 119867
119894(119878119868119881))
(33) EndIf(34) EndFor lowastend oflocal search stagelowast(35) Make sure each habitat legal based on boundary constraints(36) Use new generated habitat to replace duplicate(37) Evaluate the fitness (HSI) for trial habitat119872
119894in the new population Pop
(38) For i = 1 to NP lowastselection stagelowast(39) If119891(119872
119894) lt 119891(119867
119894)
(40) 119867119894= 119872119894
(41) EndIf(42) EndFor lowastend ofselection stagelowast(43) EndWhile(44) End
Algorithm 4 Pseudocode of MLBBO
validating the performance of the proposed algorithmThesefunctions consist of 13 standard benchmark functions [22](f1ndashf13) and 14 complex functions (F1ndashF14) from the set of testproblems listed in IEEE CEC 2005 [23] All the test functionsused in our experiments are high-dimensional functionswithchangeable dimension which are only briefly summarized inTable 1 A more detailed description of each function is givenin Table 8 (see also [22 23])
For fair comparison we have chosen a reasonable set ofparameter values For all experimental simulations in this
paper we will use the following parameters setting unless achange is mentioned
(1) population size NP = 100(2) maximum immigration rate I =1(3) maximum emigration rate E =1(4) mutation probability119898max = 0001(5) value to reach (VTR) = 10minus6 except for f7 F6ndashF14 of
VTR = 10minus2
6 Journal of Applied Mathematics
Table 1 Test unctions used in our experimental tests
Function name Dimension Search Space Optimaf1 Sphere n [minus100 100] 0f2 Schwefelrsquos Problem 222 n [10 10] 0f3 Schwefelrsquos Problem 12 n [minus100 100] 0f4 Schwefelrsquos Problem 221 n [minus100 100] 0f5 Rosenbrock function n [minus30 30] 0f6 Step function n [minus100 100] 0f7 Quartic function n [minus128 128] 0f8 Schwefel function n [minus512 512] minus125695f9 Rastrigin function n [minus512 512] 0f10 Ackley function n [minus32 32] 0f11 Griewank function n [minus600 600] 0f12 Penalized function 1 n [minus50 50] 0f13 Penalized function 2 n [minus50 50] 0F1 Shifted sphere function n [minus100 100] minus450F2 Shifted Schwefelrsquos Problem 12 n [minus100 100] minus450F3 Shifted rotated high conditioned elliptic function n [minus100 100] minus450F4 Shifted Schwefelrsquos Problem 12 with noise in fitness n [minus100 100] minus450F5 Schwefelrsquos Problem 26 with global optimum on bounds n [minus30 30] minus310F6 Shifted Rosenbrockrsquos function n [minus100 100] 390F7 Shifted rotated Griewankrsquos function without bounds n Initialize in [0 600] minus180F8 Shifted rotated Ackleyrsquos function with global optimum on bounds n [minus32 32] minus140F9 Shifted Rastriginrsquos function n [minus512 512] minus330F10 Shifted Rotated Rastriginrsquos function n [minus512 512] minus330F11 Shifted Rotated Weierstrass function n [minus600 600] 90F12 Schwefelrsquos Problem 213 n [minus100 100] minus460F13 Expanded extended Griewankrsquos plus Rosenbrockrsquos function (F8F2) n [minus50 50] minus130F14 Shifted rotated expanded scafferrsquos F6 n [minus100 100] minus300
(6) maximum number of fitness function evaluations(MaxNFFEs) 150000 for f1 f6 f10 f12 and f13200000 for f2 and f11 300000 for f7 f8 f9 F1ndashF14and 500000 for f3 f4 and f5
In our experimental tests all the functions are optimizedover 30 independent runs All experimental tests are carriedout on a computer with 186GHz Intel core Processor 1 GB ofRAM and Window XP operation system in Matlab software76
42 Effect of the Strength Alpha and Rate 119901119897on the Perfor-
mance of MLBBO In this section we investigate the impactof local search strength alpha and search probability 119901
119897on
the proposed algorithm In order to analyze the performancemore scientific eight functions are chosen from Table 1Theyare Sphere Schwefel 12 Schwefel Penalized2 F1 F3 F10 and
F13 which belong to standard unimodal function standardmultimodal function shifted function shifted rotated func-tion and expanded function respectively Search strengthalpha are 03 05 08 and 11 Search rates 119901
119897are 0 02 04
06 08 and 1 It is important to note that 119901119897equal to 0
means that local search mechanism is not conducted Hencein our experiments alpha and 119901
119897are tested according to
24 situations MLBBO algorithm executes 30 independentruns on each function and convergence figures of averagefitness values (which are function error values) are plotted inFigure 1
From Figure 1 we can observe that different search rate 119901119897
have different results When 119901119897is 0 the results are worse than
the others which means that local search mechanism canimprove the performance of proposed algorithm When 119901
119897is
04 we obtain a faster convergence speed and better resultson the Sphere function For the other seven test functions
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
4 Journal of Applied Mathematics
(1) Begin(2) For i = 1 to NP(3) Compute the mutation probability p(4) Select variable119867
119894with probability p
(5) If 119867119894is selected
(6) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(7) Endif(8) Endfor(9) End
Algorithm 3 Pseudocode of Cauchy mutation operator
In order to enhance exploration ability we modifymigration operator described in Algorithm 1 by combiningformula 119867
119894(SIV) = 119867
119895(SIV) and formula (4) The modified
migration operator is described in Algorithm 2 We can see from Algorithm 2 that immigration habitats
not only accept information from best habitat but also fromBBObest2 On the one hand information fromHSI habitatsmake the operator maintain power exploitation ability Onthe other hand information from BBObest2 can enhancethe exploration ability From this we know that this modifiedmigration operator can preserve the exploitation ability andenhance the exploration ability simultaneously
32 Mutation with Mutation Operator Cataclysmic eventscan drastically change the HSI of a natural habitat andcause species count to differ from its equilibrium value Thisphenomenon can be modeled in BBO as SIV mutation andspecies count probabilities are used to determine mutationrates
In order to enhance the exploration ability of BBO amodified mutation operator Cauchy mutation operator isintegrated into BBO The probability density function ofCauchy distribution [9] is
119891119905 (119909) =
1
120587
119905
1199052 + 1199092 (6)
where 119909 isin 119877 and 119905 gt 0 is a scale parameterTheCauchymutation (t = 1) in this paper can be described
as
119867119894 (SIV) = 119867119894 (SIV) + 120575 (1) (7)
where 120575(1) is a randomness Cauchy value when parametert equals 1 In this paper we use the Cauchy distribution toupdate the solution which can be described in Algorithm 3
33 Local Search Mechanism As Simon has pointed out theconvergence speed is an obstacle for real application Sofinding mechanisms to speed up BBO could be an importantarea for further researchTherefore a local searchmechanismis proposed to accelerate the convergence speed of BBOIn order to remedy the insufficiency of modified migration
operator which mainly updates worse habitats selected byimmigration rate the local search mechanism is designedmainly for better habitats in the first half population
For each habitat119867119894in the first half population define
119867119894 (SIV) =
119867119894 (SIV) + 119886119897119901ℎ119886 lowast (119867119896 (SIV) minus 119867119894 (SIV))
119903119886119899119889 (0 1) lt 119901119897
119867119894 (SIV) else
(8)
where 119867119896denotes a habitat randomly selected from popula-
tion 119901119897denotes the local search probability and alpha is the
strength of local search
34 Selection Operator Selection operator is used to selectbetter habitats preserving in next generationwhich can impelthemetabolismof population and retain goodhabitats In thispaper greedy selection operator is used In other words theoriginal habitat119867
119894will be replaced by its corresponding trial
habitat119872119894if the HSI of the trial habitat119872
119894is better than that
of its original habitat119867119894
119867119894= 119867119894
if 119891 (119867119894) lt 119891 (119872
119894)
119872119894
else(9)
35 Main Procedure of MLBBO Exploration ability is mainlya contributory factor to step down the performance ofBBO algorithm In order to improve the performance ofBBO a modified BBO algorithm called MLBBO is pro-posed for global optimization problems in the continu-ous domain In MLBBO the original migration opera-tor and mutation operator are modified to enhance theexploration ability Moreover local search mechanism andselection operator are integrated into MLBBO to supple-ment with migration operator The pseudo-code of MLBBOis described in Algorithm 4 where NP is the populationsize
4 Experimental Study
41 Experimental Setting In this section several experimen-tal tests are carried out on 27 benchmark functions for
Journal of Applied Mathematics 5
(1) Begin(2) Initialize the population Pop with NP habitats randomly(3) Evaluate the fitness (HSI) for each Habitat119867
119894in Pop
(4) While the halting criteria are not satisfied do(5) Sort all the habitats from best to worst in line with their fitness values(6) Map the fitness to the number of species count S for each habitat(7) Calculate the immigration rate and emigration rate according to sinusoidal migration model(8) For i = 1 to NP lowastmigration stagelowast(9) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(10) If 119867119894is selected
(11) For j = 1 to NP(12) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 sdot sdot sdot 119873119875(13) Select119867
119895according to emigration rate 120583
119895
(14) If rand(01) lt 120583119895
(15) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(16) Else(17) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(18) EndIf(19) EndFor(20) EndIf(21) EndFor lowastend ofmigration stagelowast(22) For i = 1 to NP lowastmutation stagelowast(23) Compute the mutation probability 119901
119894
(24) Select variable119867119894with probability 119901
119894
(25) If 119867119894is selected
(26) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(27) Endif(28) EndFor lowastend ofmutation stagelowast(29) For i = 1 to NP2 lowastlocal search stagelowast(30) If rand(01) lt 119901
119897
(31) Select a habitat119867119896randomly
(32) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + alpha lowast (119867
119896(119878119868119881) minus 119867
119894(119878119868119881))
(33) EndIf(34) EndFor lowastend oflocal search stagelowast(35) Make sure each habitat legal based on boundary constraints(36) Use new generated habitat to replace duplicate(37) Evaluate the fitness (HSI) for trial habitat119872
119894in the new population Pop
(38) For i = 1 to NP lowastselection stagelowast(39) If119891(119872
119894) lt 119891(119867
119894)
(40) 119867119894= 119872119894
(41) EndIf(42) EndFor lowastend ofselection stagelowast(43) EndWhile(44) End
Algorithm 4 Pseudocode of MLBBO
validating the performance of the proposed algorithmThesefunctions consist of 13 standard benchmark functions [22](f1ndashf13) and 14 complex functions (F1ndashF14) from the set of testproblems listed in IEEE CEC 2005 [23] All the test functionsused in our experiments are high-dimensional functionswithchangeable dimension which are only briefly summarized inTable 1 A more detailed description of each function is givenin Table 8 (see also [22 23])
For fair comparison we have chosen a reasonable set ofparameter values For all experimental simulations in this
paper we will use the following parameters setting unless achange is mentioned
(1) population size NP = 100(2) maximum immigration rate I =1(3) maximum emigration rate E =1(4) mutation probability119898max = 0001(5) value to reach (VTR) = 10minus6 except for f7 F6ndashF14 of
VTR = 10minus2
6 Journal of Applied Mathematics
Table 1 Test unctions used in our experimental tests
Function name Dimension Search Space Optimaf1 Sphere n [minus100 100] 0f2 Schwefelrsquos Problem 222 n [10 10] 0f3 Schwefelrsquos Problem 12 n [minus100 100] 0f4 Schwefelrsquos Problem 221 n [minus100 100] 0f5 Rosenbrock function n [minus30 30] 0f6 Step function n [minus100 100] 0f7 Quartic function n [minus128 128] 0f8 Schwefel function n [minus512 512] minus125695f9 Rastrigin function n [minus512 512] 0f10 Ackley function n [minus32 32] 0f11 Griewank function n [minus600 600] 0f12 Penalized function 1 n [minus50 50] 0f13 Penalized function 2 n [minus50 50] 0F1 Shifted sphere function n [minus100 100] minus450F2 Shifted Schwefelrsquos Problem 12 n [minus100 100] minus450F3 Shifted rotated high conditioned elliptic function n [minus100 100] minus450F4 Shifted Schwefelrsquos Problem 12 with noise in fitness n [minus100 100] minus450F5 Schwefelrsquos Problem 26 with global optimum on bounds n [minus30 30] minus310F6 Shifted Rosenbrockrsquos function n [minus100 100] 390F7 Shifted rotated Griewankrsquos function without bounds n Initialize in [0 600] minus180F8 Shifted rotated Ackleyrsquos function with global optimum on bounds n [minus32 32] minus140F9 Shifted Rastriginrsquos function n [minus512 512] minus330F10 Shifted Rotated Rastriginrsquos function n [minus512 512] minus330F11 Shifted Rotated Weierstrass function n [minus600 600] 90F12 Schwefelrsquos Problem 213 n [minus100 100] minus460F13 Expanded extended Griewankrsquos plus Rosenbrockrsquos function (F8F2) n [minus50 50] minus130F14 Shifted rotated expanded scafferrsquos F6 n [minus100 100] minus300
(6) maximum number of fitness function evaluations(MaxNFFEs) 150000 for f1 f6 f10 f12 and f13200000 for f2 and f11 300000 for f7 f8 f9 F1ndashF14and 500000 for f3 f4 and f5
In our experimental tests all the functions are optimizedover 30 independent runs All experimental tests are carriedout on a computer with 186GHz Intel core Processor 1 GB ofRAM and Window XP operation system in Matlab software76
42 Effect of the Strength Alpha and Rate 119901119897on the Perfor-
mance of MLBBO In this section we investigate the impactof local search strength alpha and search probability 119901
119897on
the proposed algorithm In order to analyze the performancemore scientific eight functions are chosen from Table 1Theyare Sphere Schwefel 12 Schwefel Penalized2 F1 F3 F10 and
F13 which belong to standard unimodal function standardmultimodal function shifted function shifted rotated func-tion and expanded function respectively Search strengthalpha are 03 05 08 and 11 Search rates 119901
119897are 0 02 04
06 08 and 1 It is important to note that 119901119897equal to 0
means that local search mechanism is not conducted Hencein our experiments alpha and 119901
119897are tested according to
24 situations MLBBO algorithm executes 30 independentruns on each function and convergence figures of averagefitness values (which are function error values) are plotted inFigure 1
From Figure 1 we can observe that different search rate 119901119897
have different results When 119901119897is 0 the results are worse than
the others which means that local search mechanism canimprove the performance of proposed algorithm When 119901
119897is
04 we obtain a faster convergence speed and better resultson the Sphere function For the other seven test functions
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 5
(1) Begin(2) Initialize the population Pop with NP habitats randomly(3) Evaluate the fitness (HSI) for each Habitat119867
119894in Pop
(4) While the halting criteria are not satisfied do(5) Sort all the habitats from best to worst in line with their fitness values(6) Map the fitness to the number of species count S for each habitat(7) Calculate the immigration rate and emigration rate according to sinusoidal migration model(8) For i = 1 to NP lowastmigration stagelowast(9) Select119867
119894according to immigration rate 120582
119894using sinusoidal migration model
(10) If 119867119894is selected
(11) For j = 1 to NP(12) Generate fourmutually different integers 1199031 1199032 1199033 and 1199034 in 1 sdot sdot sdot 119873119875(13) Select119867
119895according to emigration rate 120583
119895
(14) If rand(01) lt 120583119895
(15) 119867119894(119878119868119881) = 119867
119895(119878119868119881)
(16) Else(17) 119867
119894(119878119868119881) = 119867best(119878119868119881) + 119865 sdot (1198671199031(119878119868119881) minus 1198671199032(119878119868119881)) + 119865 sdot (1198671199033(119878119868119881) minus 1198671199034(119878119868119881))
(18) EndIf(19) EndFor(20) EndIf(21) EndFor lowastend ofmigration stagelowast(22) For i = 1 to NP lowastmutation stagelowast(23) Compute the mutation probability 119901
119894
(24) Select variable119867119894with probability 119901
119894
(25) If 119867119894is selected
(26) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + 120575(1)
(27) Endif(28) EndFor lowastend ofmutation stagelowast(29) For i = 1 to NP2 lowastlocal search stagelowast(30) If rand(01) lt 119901
119897
(31) Select a habitat119867119896randomly
(32) 119867119894(119878119868119881) = 119867
119894(119878119868119881) + alpha lowast (119867
119896(119878119868119881) minus 119867
119894(119878119868119881))
(33) EndIf(34) EndFor lowastend oflocal search stagelowast(35) Make sure each habitat legal based on boundary constraints(36) Use new generated habitat to replace duplicate(37) Evaluate the fitness (HSI) for trial habitat119872
119894in the new population Pop
(38) For i = 1 to NP lowastselection stagelowast(39) If119891(119872
119894) lt 119891(119867
119894)
(40) 119867119894= 119872119894
(41) EndIf(42) EndFor lowastend ofselection stagelowast(43) EndWhile(44) End
Algorithm 4 Pseudocode of MLBBO
validating the performance of the proposed algorithmThesefunctions consist of 13 standard benchmark functions [22](f1ndashf13) and 14 complex functions (F1ndashF14) from the set of testproblems listed in IEEE CEC 2005 [23] All the test functionsused in our experiments are high-dimensional functionswithchangeable dimension which are only briefly summarized inTable 1 A more detailed description of each function is givenin Table 8 (see also [22 23])
For fair comparison we have chosen a reasonable set ofparameter values For all experimental simulations in this
paper we will use the following parameters setting unless achange is mentioned
(1) population size NP = 100(2) maximum immigration rate I =1(3) maximum emigration rate E =1(4) mutation probability119898max = 0001(5) value to reach (VTR) = 10minus6 except for f7 F6ndashF14 of
VTR = 10minus2
6 Journal of Applied Mathematics
Table 1 Test unctions used in our experimental tests
Function name Dimension Search Space Optimaf1 Sphere n [minus100 100] 0f2 Schwefelrsquos Problem 222 n [10 10] 0f3 Schwefelrsquos Problem 12 n [minus100 100] 0f4 Schwefelrsquos Problem 221 n [minus100 100] 0f5 Rosenbrock function n [minus30 30] 0f6 Step function n [minus100 100] 0f7 Quartic function n [minus128 128] 0f8 Schwefel function n [minus512 512] minus125695f9 Rastrigin function n [minus512 512] 0f10 Ackley function n [minus32 32] 0f11 Griewank function n [minus600 600] 0f12 Penalized function 1 n [minus50 50] 0f13 Penalized function 2 n [minus50 50] 0F1 Shifted sphere function n [minus100 100] minus450F2 Shifted Schwefelrsquos Problem 12 n [minus100 100] minus450F3 Shifted rotated high conditioned elliptic function n [minus100 100] minus450F4 Shifted Schwefelrsquos Problem 12 with noise in fitness n [minus100 100] minus450F5 Schwefelrsquos Problem 26 with global optimum on bounds n [minus30 30] minus310F6 Shifted Rosenbrockrsquos function n [minus100 100] 390F7 Shifted rotated Griewankrsquos function without bounds n Initialize in [0 600] minus180F8 Shifted rotated Ackleyrsquos function with global optimum on bounds n [minus32 32] minus140F9 Shifted Rastriginrsquos function n [minus512 512] minus330F10 Shifted Rotated Rastriginrsquos function n [minus512 512] minus330F11 Shifted Rotated Weierstrass function n [minus600 600] 90F12 Schwefelrsquos Problem 213 n [minus100 100] minus460F13 Expanded extended Griewankrsquos plus Rosenbrockrsquos function (F8F2) n [minus50 50] minus130F14 Shifted rotated expanded scafferrsquos F6 n [minus100 100] minus300
(6) maximum number of fitness function evaluations(MaxNFFEs) 150000 for f1 f6 f10 f12 and f13200000 for f2 and f11 300000 for f7 f8 f9 F1ndashF14and 500000 for f3 f4 and f5
In our experimental tests all the functions are optimizedover 30 independent runs All experimental tests are carriedout on a computer with 186GHz Intel core Processor 1 GB ofRAM and Window XP operation system in Matlab software76
42 Effect of the Strength Alpha and Rate 119901119897on the Perfor-
mance of MLBBO In this section we investigate the impactof local search strength alpha and search probability 119901
119897on
the proposed algorithm In order to analyze the performancemore scientific eight functions are chosen from Table 1Theyare Sphere Schwefel 12 Schwefel Penalized2 F1 F3 F10 and
F13 which belong to standard unimodal function standardmultimodal function shifted function shifted rotated func-tion and expanded function respectively Search strengthalpha are 03 05 08 and 11 Search rates 119901
119897are 0 02 04
06 08 and 1 It is important to note that 119901119897equal to 0
means that local search mechanism is not conducted Hencein our experiments alpha and 119901
119897are tested according to
24 situations MLBBO algorithm executes 30 independentruns on each function and convergence figures of averagefitness values (which are function error values) are plotted inFigure 1
From Figure 1 we can observe that different search rate 119901119897
have different results When 119901119897is 0 the results are worse than
the others which means that local search mechanism canimprove the performance of proposed algorithm When 119901
119897is
04 we obtain a faster convergence speed and better resultson the Sphere function For the other seven test functions
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
6 Journal of Applied Mathematics
Table 1 Test unctions used in our experimental tests
Function name Dimension Search Space Optimaf1 Sphere n [minus100 100] 0f2 Schwefelrsquos Problem 222 n [10 10] 0f3 Schwefelrsquos Problem 12 n [minus100 100] 0f4 Schwefelrsquos Problem 221 n [minus100 100] 0f5 Rosenbrock function n [minus30 30] 0f6 Step function n [minus100 100] 0f7 Quartic function n [minus128 128] 0f8 Schwefel function n [minus512 512] minus125695f9 Rastrigin function n [minus512 512] 0f10 Ackley function n [minus32 32] 0f11 Griewank function n [minus600 600] 0f12 Penalized function 1 n [minus50 50] 0f13 Penalized function 2 n [minus50 50] 0F1 Shifted sphere function n [minus100 100] minus450F2 Shifted Schwefelrsquos Problem 12 n [minus100 100] minus450F3 Shifted rotated high conditioned elliptic function n [minus100 100] minus450F4 Shifted Schwefelrsquos Problem 12 with noise in fitness n [minus100 100] minus450F5 Schwefelrsquos Problem 26 with global optimum on bounds n [minus30 30] minus310F6 Shifted Rosenbrockrsquos function n [minus100 100] 390F7 Shifted rotated Griewankrsquos function without bounds n Initialize in [0 600] minus180F8 Shifted rotated Ackleyrsquos function with global optimum on bounds n [minus32 32] minus140F9 Shifted Rastriginrsquos function n [minus512 512] minus330F10 Shifted Rotated Rastriginrsquos function n [minus512 512] minus330F11 Shifted Rotated Weierstrass function n [minus600 600] 90F12 Schwefelrsquos Problem 213 n [minus100 100] minus460F13 Expanded extended Griewankrsquos plus Rosenbrockrsquos function (F8F2) n [minus50 50] minus130F14 Shifted rotated expanded scafferrsquos F6 n [minus100 100] minus300
(6) maximum number of fitness function evaluations(MaxNFFEs) 150000 for f1 f6 f10 f12 and f13200000 for f2 and f11 300000 for f7 f8 f9 F1ndashF14and 500000 for f3 f4 and f5
In our experimental tests all the functions are optimizedover 30 independent runs All experimental tests are carriedout on a computer with 186GHz Intel core Processor 1 GB ofRAM and Window XP operation system in Matlab software76
42 Effect of the Strength Alpha and Rate 119901119897on the Perfor-
mance of MLBBO In this section we investigate the impactof local search strength alpha and search probability 119901
119897on
the proposed algorithm In order to analyze the performancemore scientific eight functions are chosen from Table 1Theyare Sphere Schwefel 12 Schwefel Penalized2 F1 F3 F10 and
F13 which belong to standard unimodal function standardmultimodal function shifted function shifted rotated func-tion and expanded function respectively Search strengthalpha are 03 05 08 and 11 Search rates 119901
119897are 0 02 04
06 08 and 1 It is important to note that 119901119897equal to 0
means that local search mechanism is not conducted Hencein our experiments alpha and 119901
119897are tested according to
24 situations MLBBO algorithm executes 30 independentruns on each function and convergence figures of averagefitness values (which are function error values) are plotted inFigure 1
From Figure 1 we can observe that different search rate 119901119897
have different results When 119901119897is 0 the results are worse than
the others which means that local search mechanism canimprove the performance of proposed algorithm When 119901
119897is
04 we obtain a faster convergence speed and better resultson the Sphere function For the other seven test functions
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 7
Table 2 Comparisons with BBO and DE
DE BBO MLBBO
Mean (std) MeanFEs SR Mean (Std) MeanFEs SR Mean (std) MeanFEs SR
f1 272119864 minus 20 (209119864 minus 20)+447119864 + 04 30 323119864 minus 01 (118119864 minus 01)
+ NaN 0 278119864 minus 31 (325119864 minus 31) 283119864 + 04 30
f2 256119864 minus 12 (309119864 minus 12)+106119864 + 05 30 484119864 minus 01 (906119864 minus 02)
+ NaN 0 143119864 minus 21 (711119864 minus 22) 574119864 + 04 30
f3 231119864 minus 19 (237119864 minus 19)+151119864 + 05 30 126119864 + 02 (519119864 + 01)
+ NaN 0 190119864 minus 20 (303119864 minus 20) 140119864 + 05 30
f4 192119864 minus 10 (344119864 minus 10)sim261119864 + 05 30 162119864 + 00 (471119864 minus 01)
+ NaN 0 449119864 minus 08 (125119864 minus 07) 349119864 + 05 30
f5 105119864 minus 28 (892119864 minus 29)minus162119864 + 05 30 785119864 + 01 (351119864 + 01)
+ NaN 0 334119864 minus 21 (908119864 minus 21) 225119864 + 05 30
f6 867119864 minus 01 (120119864 + 00)+251119864 + 04 14 177119864 + 00 (117119864 + 00)
+119119864 + 05 3 000119864 + 00 (000119864 + 00) 114119864 + 04 30
f7 565119864 minus 03 (210119864 minus 03)+144119864 + 05 29 364119864 minus 04 (277119864 minus 04)
minus221119864 + 04 30 223119864 minus 03 (991119864 minus 04) 661119864 + 04 30
f8 729119864 + 03 (232119864 + 02)+ NaN 0 253119864 minus 01 (882119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 553119864 + 04 30
f9 176119864 + 02 (101119864 + 01)+ NaN 0 356119864 minus 02 (152119864 minus 02)
+ NaN 0 000119864 + 00 (000119864 + 00) 916119864 + 04 30
f10 577119864 minus 07 (316119864 minus 06)sim787119864 + 04 29 206119864 minus 01 (494119864 minus 02)
+ NaN 0 610119864 minus 15 (114119864 minus 15) 505119864 + 04 30
f11 657119864 minus 03 (625119864 minus 03)+454119864 + 04 12 281119864 minus 01 (976119864 minus 02)
+ NaN 0 287119864 minus 03 (502119864 minus 03) 333119864 + 04 22
f12 138119864 minus 02 (450119864 minus 02)sim537119864 + 04 26 198119864 minus 03 (204119864 minus 03)
+ NaN 0 588119864 minus 28 (322119864 minus 27) 257119864 + 04 30
f13 190119864 minus 20 (593119864 minus 20)sim465119864 + 04 30 209119864 minus 02 (994119864 minus 03)
+ NaN 0 509119864 minus 32 (313119864 minus 32) 271119864 + 04 30
F1 559119864 minus 28 (206119864 minus 28)+458119864 + 04 30 723119864 minus 02 (327119864 minus 02)
+ NaN 0 303119864 minus 29 (697119864 minus 29) 290119864 + 04 30
F2 759119864 minus 12 (842119864 minus 12)+172119864 + 05 30 355119864 + 03 (135119864 + 03)
+ NaN 0 197119864 minus 12 (255119864 minus 12) 158119864 + 05 30
F3 139119864 + 05 (106119864 + 05)+ NaN 0 114119864 + 07 (490119864 + 06)
+ NaN 0 926119864 + 04 (490119864 + 04) NaN 0
F4 710119864 minus 05 (161119864 minus 04)sim285119864 + 05 1 139119864 + 04 (456119864 + 03)
+ NaN 0 700119864 minus 05 (187119864 minus 04) 290119864 + 05 5
F5 531119864 + 01 (137119864 + 02)minus NaN 0 581119864 + 03 (997119864 + 02)
+ NaN 0 834119864 + 02 (381119864 + 02) NaN 0
F6 158119864 minus 13 (782119864 minus 13)sim146119864 + 05 30 223119864 + 02 (803119864 + 01)
+ NaN 0 211119864 minus 09 (668119864 minus 09) 185119864 + 05 30
F7 126119864 minus 02 (126119864 minus 02)sim493119864 + 04 15 113119864 + 00 (773119864 minus 02)
+ NaN 0 159119864 minus 02 (116119864 minus 02) 480119864 + 04 12
F8 209119864 + 01 (520119864 minus 02)sim NaN 0 208119864 + 01 (106119864 minus 01)
minus NaN 0 209119864 + 01 (707119864 minus 02) NaN 0
F9 185119864 + 02 (175119864 + 01)+ NaN 0 397119864 minus 02 (204119864 minus 02)
+ NaN 0 592119864 minus 17 (324119864 minus 16) 771119864 + 04 30
F10 205119864 + 02 (196119864 + 01)+ NaN 0 526119864 + 01 (119119864 + 01)
+ NaN 0 347119864 + 01 (124119864 + 01) NaN 0
F11 395119864 + 01 (103119864 + 00)+ NaN 0 309119864 + 01 (305119864 + 00)
+ NaN 0 172119864 + 01 (658119864 + 00) NaN 0
F12 119119864 + 03 (232119864 + 03)sim231119864 + 05 3 141119864 + 04 (984119864 + 03)
+ NaN 0 207119864 + 03 (292119864 + 03) NaN 0
F13 159119864 + 01 (112119864 + 00)+ NaN 0 109119864 + 00 (234119864 minus 01)
minus NaN 0 303119864 + 00 (144119864 minus 01) NaN 0
F14 134119864 + 01 (151119864 minus 01)+ NaN 0 131119864 + 01 (383119864 minus 01)
+ NaN 0 127119864 + 01 (240119864 minus 01) NaN 0
better 16 24
similar 9 0
worse 2 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
better results are obtained when search rate 119901119897is 02 and 04
Moreover the performance of MLBBO is most sensitive tosearch rate 119901
119897on function F8
From Figure 1 we can also see that search strength alphahas little effect on the results as a whole By careful looking
search strength alpha has an effect on the results when searchrate 119901
119897is large especially when 119901
119897equals 08 or 1
For considering opinions of both sides the searchstrength alpha and search rate 119901
119897are set as 08 and 02
respectively
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
8 Journal of Applied Mathematics
alpha = 03alpha = 05
alpha = 08alpha = 11
alpha = 03alpha = 05
alpha = 08alpha = 11
10minus45
10minus40
10minus35
10minus30
10minus25
10minus20
10minus15
Fitn
ess v
alue
s (lo
g)
Sphere
10minus15
10minus10
10minus5
100
105
Schwefel
Local search rate
Fitn
ess v
alue
s (lo
g)
10minus30
10minus25
10minus20
10minus15
10minus10
Schwefel2
10minus35
10minus30
10minus25
10minus20
10minus15 Penalized2
10minus28
10minus27
10minus26
10minus25
F1
104
105
106 F3
0 02 04 06 08 1Local search rate
0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
Local search rate0 02 04 06 08 1
101321
101322
1005
1006
1007
1008
1009
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)Fi
tnes
s val
ues (
log)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
Fitn
ess v
alue
s (lo
g)
F8 F13
Figure 1 Results of MLBBO on eight test functions with different alpha and 119901119897
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 9
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ssSphere
0 05 1 15 2times10
5
10minus40
10minus20
100
1020
1040 Schwefel3
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus10
10minus5
100
105
Schwefel4
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 4 5times10
5
10minus30
10minus20
10minus10
100
1010 Rosenbrock
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15times10
4
10minus2
100
102
104
106 Step
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
100
Quartic
Number of fitness function evaluations
Best
fitne
ss0 5 10 15
times104
10minus20
100
Penalized2
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus30
10minus20
10minus10
100
1010 F1
Number of fitness function evaluationsBe
st fit
ness
0 05 1 15 2 25 3times10
5
104
106
108
1010 F3
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus5
100
105
F4
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
101
102
103
104
105 F5
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25times10
5
10minus2
100
102
F7
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus15
10minus10
10minus5
100
105 Schwefel
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
F8
Number of fitness function evaluations
101319
101322
101325
101328
Best
fitne
ss
0 05 1 15 2 25 3times10
5
10minus20
10minus10
100
1010 F9
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3times10
5
1013
1014
1015
1016
F11
Number of fitness function evaluations
Best
fitne
ss
MLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
0 05 1 15 2 25times10
5
100
102
F13
Best
fitne
ss
Number of fitness function evaluationsMLBBODE
BBO
Figure 2 Convergence curves of mean fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8F9 F11 F13 and F14
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
10 Journal of Applied Mathematics
0
02
04
06
Schwefel3
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
Schwefel4
MLBBO DE BBO
Best
fitne
ss
0
50
100
150Rosenbrock
MLBBO DE BBO
Best
fitne
ss
0
1
2
3
4
5Step
MLBBO DE BBO
Best
fitne
ss
02468
10times10
minus3 Quartic
MLBBO DE BBO
Best
fitne
ss0
2000
4000
6000
8000 Schwefel
MLBBO DE BBO
Best
fitne
ss
0001002003004005
Penalized2
MLBBO DE BBO
Best
fitne
ss
0
005
01
015
F1
MLBBO DE BBOBe
st fit
ness
005
115
225times10
7 F3
MLBBO DE BBO
Best
fitne
ss
005
115
225times10
4 F4
MLBBO DE BBO
Best
fitne
ss
0
2000
4000
6000
8000F5
MLBBO DE BBO
Best
fitne
ss
0
05
1
F7
MLBBO DE BBO
Best
fitne
ss
206
207
208
209
21
F8
MLBBO DE BBO
Best
fitne
ss
0
50
100
150
200F9
MLBBO DE BBO
Best
fitne
ss
0
02
04
06Be
st fit
ness
Sphere
MLBBO DE BBO
Best
fitne
ss
10
20
30
40F11
MLBBO DE BBO
Best
fitne
ss
0
5
10
15
F13
MLBBO DE BBO
Best
fitne
ss
12
125
13
135
F14
MLBBO DE BBO
Best
fitne
ss
Figure 3 ANOVA tests of fitness values of MLBBO BBO and DE for functions f1 f2 f3 f5 f6 f7 f8 f13 F1 F3 F4 F5 F7 F8 F9 F11 F13and F14
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 11
Table 3 Comparisons with improved BBO algorithms
OBBO PBBO MOBBO RCBBO MLBBOMean (std) Mean (Std) Mean (std) Mean (std) Mean (std)
f1 688119864 minus 05 (258119864 minus 05)+223119864 minus 11 (277119864 minus 12)
+170119864 minus 09 (930119864 minus 09)
sim226119864 minus 03 (106119864 minus 03)
+211119864 minus 31 (125119864 minus 31)
f2 276119864 minus 02 (480119864 minus 03)+347119864 minus 06 (267119864 minus 07)
+274119864 minus 05 (148119864 minus 04)
sim896119864 minus 02 (158119864 minus 02)
+158119864 minus 21 (725119864 minus 22)
f3 629119864 minus 01 (423119864 minus 01)+206119864 minus 02 (170119864 minus 02)
+325119864 minus 02 (450119864 minus 02)
+219119864 + 01 (934119864 + 00)
+142119864 minus 20 (170119864 minus 20)
f4 147119864 minus 02 (318119864 minus 03)+120119864 minus 02 (330119864 minus 03)
+243119864 minus 02 (674119864 minus 03)
+120119864 + 00 (125119864 + 00)
+528119864 minus 08 (102119864 minus 07)
f5 365119864 + 01 (220119864 + 01)+360119864 + 01 (340119864 + 01)
+632119864 + 01 (833119864 + 01)
+425119864 + 01 (374119864 + 01)
+534119864 minus 21 (110119864 minus 20)
f6 000119864 + 00 (000119864 + 00)sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
f7 402119864 minus 04 (348119864 minus 04)minus286119864 minus 04 (268119864 minus 04)
minus979119864 minus 04 (800119864 minus 04)
minus421119864 minus 04 (381119864 minus 04)
minus233119864 minus 03 (104119864 minus 03)
f8 240119864 + 02 (182119864 + 02)+170119864 minus 11 (226119864 minus 12)
+155119864 minus 11 (850119864 minus 11)
sim817119864 minus 05 (315119864 minus 05)
+000119864 + 00 (000119864 + 00)
f9 559119864 minus 03 (207119864 minus 03)+280119864 minus 07 (137119864 minus 06)
sim545119864 minus 05 (299119864 minus 04)
sim138119864 minus 02 (511119864 minus 03)
+000119864 + 00 (000119864 + 00)
f10 738119864 minus 03 (188119864 minus 03)+208119864 minus 06 (539119864 minus 06)
+974119864 minus 08 (597119864 minus 09)
+294119864 minus 02 (663119864 minus 03)
+610119864 minus 15 (649119864 minus 16)
f11 268119864 minus 02 (249119864 minus 02)+498119864 minus 03 (129119864 minus 02)
sim508119864 minus 03 (105119864 minus 02)
sim792119864 minus 02 (436119864 minus 02)
+173119864 minus 03 (400119864 minus 03)
f12 182119864 minus 06 (226119864 minus 06)+671119864 minus 11 (318119864 minus 10)
sim346119864 minus 03 (189119864 minus 02)
sim339119864 minus 05 (347119864 minus 05)
+235119864 minus 32 (398119864 minus 32)
f13 204119864 minus 05 (199119864 minus 05)+165119864 minus 09 (799119864 minus 09)
sim633119864 minus 13 (344119864 minus 12)
sim643119864 minus 04 (966119864 minus 04)
+564119864 minus 32 (338119864 minus 32)
F1 140119864 minus 05 (826119864 minus 06)+267119864 minus 11 (969119864 minus 11)
sim116119864 minus 07 (634119864 minus 07)
sim350119864 minus 04 (925119864 minus 05)
+202119864 minus 29 (616119864 minus 29)
F2 470119864 + 01 (864119864 + 00)+292119864 + 00 (202119864 + 00)
+266119864 + 00 (216119864 + 00)
+872119864 + 02 (443119864 + 02)
+137119864 minus 12 (153119864 minus 12)
F3 156119864 + 06 (345119864 + 05)+210119864 + 06 (686119864 + 05)
+237119864 + 06 (989119864 + 05)
+446119864 + 06 (158119864 + 06)
+902119864 + 04 (554119864 + 04)
F4 318119864 + 03 (994119864 + 02)+774119864 + 02 (386119864 + 02)
+146119864 + 01 (183119864 + 01)
+217119864 + 04 (850119864 + 03)
+175119864 minus 05 (292119864 minus 05)
F5 103119864 + 04 (163119864 + 03)+368119864 + 03 (627119864 + 02)
+504119864 + 03 (124119864 + 03)
+631119864 + 03 (118119864 + 03)
+800119864 + 02 (379119864 + 02)
F6 320119864 + 02 (964119864 + 02)sim871119864 + 02 (212119864 + 03)
+276119864 + 02 (846119864 + 02)
sim845119864 + 02 (269119864 + 03)
sim355119864 minus 09 (140119864 minus 08)
F7 246119864 minus 02 (152119864 minus 02)+162119864 minus 02 (145119864 minus 02)
sim226119864 minus 02 (199119864 minus 02)
+764119864 minus 01 (139119864 + 00)
+138119864 minus 02 (121119864 minus 02)
F8 208119864 + 01 (804119864 minus 02)minus209119864 + 01 (169119864 minus 01)
minus207119864 + 01 (171119864 minus 01)
minus207119864 + 01 (851119864 minus 02)
minus209119864 + 01 (548119864 minus 02)
F9 580119864 minus 03 (239119864 minus 03)+969119864 minus 07 (388119864 minus 06)
sim122119864 minus 07 (661119864 minus 07)
sim132119864 minus 02 (553119864 minus 03)
+592119864 minus 17 (324119864 minus 16)
F10 643119864 + 01 (281119864 + 01)+329119864 + 01 (880119864 + 00)
sim714119864 + 01 (206119864 + 01)
+480119864 + 01 (162119864 + 01)
+369119864 + 01 (145119864 + 01)
F11 236119864 + 01 (319119864 + 00)+209119864 + 01 (448119864 + 00)
sim270119864 + 01 (411119864 + 00)
+319119864 + 01 (390119864 + 00)
+195119864 + 01 (787119864 + 00)
F12 118119864 + 04 (866119864 + 03)+468119864 + 03 (444119864 + 03)
+504119864 + 03 (410119864 + 03)
+129119864 + 04 (894119864 + 03)
+243119864 + 03 (297119864 + 03)
F13 108119864 + 00 (196119864 minus 01)minus114119864 + 00 (207119864 minus 01)
minus109119864 + 00 (229119864 minus 01)
minus961119864 minus 01 (157119864 minus 01)
minus298119864 + 00 (162119864 minus 01)
F14 125119864 + 01 (361119864 minus 01)sim119119864 + 01 (617119864 minus 01)
minus133119864 + 01 (344119864 minus 01)
+131119864 + 01 (339119864 minus 01)
+127119864 + 01 (239119864 minus 01)
Better 21 13 13 22Similar 3 10 11 2Worse 3 4 3 3Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
43 Comparisons with BBO and DEBest2Bin For evaluat-ing the performance of algorithms we first compareMLBBOwith BBO [8] which is consistent with original BBO exceptsolution representation (real code) and DEbest2bin [7](abbreviation as DE) For fair comparison three algorithmsare conducted on 27 functions with the same parameterssetting above The average and standard deviation of fitnessvalues (which are function error values) after completionof 30 Monte Carlo simulations are summarized in Table 2The number of successful runs and average of the numberof fitness function evaluations (MeanFEs) needed in eachrun when the VTR is reached within the MaxNFFEs arealso recorded in Table 2 Furthermore results of MLBBOand BBO MLBBO and DE are compared with statisticalpaired 119905-test [24 25] which is used to determine whethertwo paired solution sets differ from each other in a significantway Convergence curves of mean fitness values and boxplotfigures under 30 independent runs are listed in Figures 2 and3 respectively
FromTable 2 it is seen thatMLBBO is significantly betterthan BBO on 24 functions out of 27 test functions andMLBBO is outperformed by BBO on the rest 3 functionsWhen compared with DE MLBBO gives significantly betterperformance on 16 functions On 9 functions there are nosignificant differences between MLBBO and DE based onthe two tail 119905-test For the rest 2 functions DE are betterthan MLBBO Moreover MLBBO can obtain the VTR overall 30 successful runs within the MaxNFFEs for 16 out of27 functions Especially for the first 13 standard benchmarkfunctionsMLBBOobtains 30 successful runs on 12 functionsexcept Griewank function where 22 successful runs areobtained by MLBBO DE and BBO obtain 30 successful runson 9 and 1 out of 27 functions respectively From Table 2we can find that MLBBO needs less or similar NFFEs thanDE and BBO on most of the successful runs which meansthatMLBBO has faster convergence speed thanDE and BBOThe convergence speed and the robustness can also show inFigures 2 and 3
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
12 Journal of Applied Mathematics
Table 4 Comparisons with hybrid algorithms
OXDE DEBBO MLBBOMean (std) MeanFEs SR Mean (std) MeanFEs SR Mean (std) MeanFEs SR
f1 260119864 minus 03 (904119864 minus 04)+ NaN 0 153119864 minus 30 (148119864 minus 30) + 47119864 + 04 30 230119864 minus 31 (274119864 minus 31) 28119864 + 04 30f2 182119864 minus 02 (477119864 minus 03)+ NaN 0 252119864 minus 24 (174119864 minus 24)minus 67119864 + 04 30 161119864 minus 21 (789119864 minus 22) 58119864 + 04 30f3 543119864 minus 01 (252119864 minus 01)+ NaN 0 260119864 minus 03 (172119864 minus 03)+ NaN 0 216119864 minus 20 (319119864 minus 20) 14119864 + 05 30f4 480119864 minus 02 (719119864 minus 02)+ NaN 0 650119864 minus 02 (223119864 minus 01)sim 25119864 + 05 18 799119864 minus 08 (138119864 minus 07) 35119864 + 05 30f5 373119864 minus 01 (768119864 minus 01)+ NaN 0 233119864 + 01 (910119864 + 00)+ NaN 0 256119864 minus 21 (621119864 minus 21) 23119864 + 05 30f6 000119864 + 00 (000119864 + 00)sim 93119864 + 04 30 000119864 + 00 (000119864 + 00)sim 22119864 + 04 30 000119864 + 00 (000119864 + 00) 11119864 + 04 30f7 664119864 minus 03 (197119864 minus 03)+ 20119864 + 05 30 469119864 minus 03 (154119864 minus 03)+ 14119864 + 05 30 213119864 minus 03 (910119864 minus 04) 67119864 + 04 30f8 133119864 minus 05 (157119864 minus 05)+ NaN 0 000119864 + 00 (000119864 + 00)sim 10119864 + 05 30 606119864 minus 14 (332119864 minus 13) 55119864 + 04 30f9 532119864 + 01 (672119864 + 00)+ NaN 0 000119864 + 00 (000119864 + 00)sim 18119864 + 05 30 000119864 + 00 (000119864 + 00) 92119864 + 04 30f10 145119864 minus 02 (275119864 minus 03)+ NaN 0 610119864 minus 15 (649119864 minus 16)sim 67119864 + 04 30 610119864 minus 15 (649119864 minus 16) 50119864 + 04 30f11 117119864 minus 04 (135119864 minus 04)minus NaN 0 000119864 + 00 (000119864 + 00)minus 50119864 + 04 30 337119864 minus 03 (464119864 minus 03) 28119864 + 04 19f12 697119864 minus 05 (320119864 minus 05)+ NaN 0 168119864 minus 31 (143119864 minus 31)sim 43119864 + 04 30 163119864 minus 25 (895119864 minus 25) 27119864 + 04 30f13 510119864 minus 04 (219119864 minus 04)+ NaN 0 237119864 minus 30 (269119864 minus 30)+ 47119864 + 04 30 498119864 minus 32 (419119864 minus 32) 27119864 + 04 30F1 157119864 minus 09 (748119864 minus 10)+ 24119864 + 05 30 000119864 + 00 (000119864 + 00)minus 48119864 + 04 30 269119864 minus 29 (698119864 minus 29) 29119864 + 04 30F2 608119864 + 02 (192119864 + 02)+ NaN 0 108119864 + 01 (107119864 + 01)+ NaN 0 160119864 minus 12 (157119864 minus 12) 16119864 + 05 30F3 826119864 + 06 (215119864 + 06)+ NaN 0 291119864 + 06 (101119864 + 06)+ NaN 0 900119864 + 04 (499119864 + 04) NaN 0F4 228119864 + 03 (746119864 + 02)+ NaN 0 149119864 + 02 (827119864 + 01)+ NaN 0 670119864 minus 05 (187119864 minus 04) 29119864 + 05 6F5 356119864 + 02 (942119864 + 01)minus NaN 0 117119864 + 03 (283119864 + 02)+ NaN 0 828119864 + 02 (340119864 + 02) NaN 0F6 203119864 + 01 (174119864 + 00)+ NaN 0 289119864 + 01 (161119864 + 01)+ NaN 0 176119864 minus 09 (658119864 minus 09) 19119864 + 05 30F7 352119864 minus 03 (107119864 minus 02)minus 25119864 + 05 27 764119864 minus 03 (542119864 minus 03)minus 13119864 + 05 29 147119864 minus 02 (147119864 minus 02) 50119864 + 04 19F8 209119864 + 01(572119864 minus 02)sim NaN 0 209119864 + 01 (484119864 minus 02)sim NaN 0 210119864 + 01 (454119864 minus 02) NaN 0F9 746119864 + 01 (103119864 + 01)+ NaN 0 000119864 + 00 (000119864 + 00)sim 15119864 + 05 30 000119864 + 00 (000119864 + 00) 78119864 + 04 30F10 187119864 + 02 (113119864 + 01)+ NaN 0 795119864 + 01 (919119864 + 00)+ NaN 0 337119864 + 01 (916119864 + 00) NaN 0F11 396119864 + 01 (953119864 minus 01)+ NaN 0 309119864 + 01 (135119864 + 00)+ NaN 0 183119864 + 01 (849119864 + 00) NaN 0F12 478119864 + 03 (449119864 + 03)+ NaN 0 425119864 + 03 (387119864 + 03)+ NaN 0 157119864 + 03 (187119864 + 03) NaN 0F13 108119864 + 01 (791119864 minus 01)+ NaN 0 278119864 + 00 (194119864 minus 01)minus NaN 0 305119864 + 00 (188119864 minus 01) NaN 0F14 134119864 + 01 (133119864 minus 01)+ NaN 0 129119864 + 01 (119119864 minus 01)+ NaN 0 128119864 + 01 (300119864 minus 01) NaN 0Better 22 14Similar 2 8Worse 3 5Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
From the above results we know that the total per-formance of MLBBO is significantly better than DE andBBO which indicates that MLBBO not only integrates theexploration ability of DE and the exploitation ability of BBOsimply but also can balance both search abilities very well
44 Comparisons with Improved BBOAlgorithms In this sec-tion the performance ofMLBBO is compared with improvedBBO algorithms (a) oppositional BBO (OBBO) [12] whichis proposed by employing opposition-based learning (OBL)alongside BBOrsquos migration rates (b) multi-operator BBO(MOBBO) [18] which is another modified BBO with multi-operator migration operator (c) perturb BBO (PBBO) [16]which is a modified BBO with perturb migration operatorand (d) real-code BBO (RCBBO) [9] For fair comparisonfive improved BBO algorithms are conducted on 27 functions
with the same parameters setting above The mean and stan-dard deviation of fitness values are summarized in Table 3Further the results between MLBBO and OBBO MLBBOand PBBO MLBBO andMOBBO and MLBBO and RCBBOare compared using statistical paired 119905-test Convergencecurves of mean fitness values under 30 independent runs areplotted in Figure 4
From Table 3 we can see that MLBBO outperformsOBBOon 21 out of 27 test functionswhereasOBBO surpassesMLBBO on 3 functions MLBBO obtains similar fitnessvalues with OBBO on 3 functions based on two tail 119905-testSimilar results can be obtained when making a comparisonbetween MLBBO and RCBBO When compared with PBBOandMOBBOMLBBO is significantly better than both PBBOandMOBBO on 13 functions MLBBO obtains similar resultswith PBBO andMOBBO on 10 and 11 functions respectivelyPBBO surpasses MLBBO only on 4 functions which are
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 13
Quartic function shifted rotatedAckleyrsquos function expandedextended Griewankrsquos plus Rosenbrockrsquos Function and shiftedrotated expanded Scafferrsquos F6MOBBOoutperformsMLBBOonly on 3 functions which are Quartic function shiftedrotatedAckleyrsquos function and expanded extendedGriewankrsquosplus Rosenbrockrsquos function Furthermore the total perfor-mance of MLBBO is better than PBBO and MOBBO andmuch better than OBBO and RCBBO
Furthermore Figure 4 shows similar results with Table 3We can see that the convergence speed of PBBO an MOBBOare faster than the others in preliminary stages owingto their powerful exploitation ability but the convergencespeed decreases in the next iteration However MLBBOcan converge to satisfactory solutions at equilibrium rapidlyspeed From the comparison we know that operators used inMLBBO can enhance exploration ability
45 Comparisons with Hybrid Algorithms In order to vali-date the performance of the proposed algorithms (MLBBO)further we compare the proposed algorithm with hybridalgorithmsOXDE [26] andDEBBO [15]Themean standarddeviation of fitness values of 27 test functions are listed inTable 4The averageNFFEs and successful runs are also listedin Table 4 Convergence curves of mean fitness values under30 independent runs are plotted in Figure 5
From Table 4 we can see that results of MLBBO aresignificantly better than those of OXDE in 22 out of 27high-dimensional functions while the results of OXDEoutperform those of MLBBO on only 3 functions Whencompared with DEBBO MLBBO outperforms DEBBOon 14 functions out of 27 test functions whereas DEBBOsurpasses MLBBO on 5 functions For the rest 8 functionsthere is no significant difference based on two tails 119905-testMLBBO DEBBO and OXDE obtain the VTR over all30 successful runs within the MaxNFFEs for 16 12 and 3test functions respectively From this we can come to aconclusion that MLBBO is more robust than DEBBO andOXDE which can be proved in Figure 5
Further more the average NFFEsMLBBO needed for thesuccessful run is less than DEBBO onmost functions whichcan also be indicate in Figure 5
46 Effect of Dimensionality In order to investigate theinfluence of scalability on the performance of MLBBO wecarry out a scalability study with DE original BBO DEBBO[15] and aBBOmDE [14] for scalable functions In the exper-imental study the first 13 standard benchmark functions arestudied at Dim = 10 50 100 and 200 and the other testfunctions are studied at Dim = 10 50 and 100 For Dim =10 the population size is equal to 50 and the population sizeis set to 100 for other dimensions in this paper MaxNFFEsis set to Dim lowast 10000 All the functions are optimized over30 independent runs in this paper The average and standarddeviation of fitness values of four compared algorithms aresummarized in Table 5 for standard test functions f1ndashf13 andTable 6 for CEC 2005 test functions F1ndashF14 The convergencecurves of mean fitness values under 30 independent runs arelisted in Figure 6 Furthermore the results of statistical paired
119905-test betweenMLBBO and others are also list in Tables 5 and6
It is to note that the results of DEBBO [15] for standardfunctions f1ndashf13 are taken from corresponding paper under50 independent runs The results of aBBOmDE [14] for allthe functions are taken from the paper directly which arecarried out with different population size (for Dim = 10 thepopulation size equals 30 and for other Dimensions it sets toDim) under 50 independently runs
For standard functions f1ndashf13 we can see from Table 5that the overall results are decreasing for five comparedalgorithms since increasing the problem dimensions leadsto algorithms that are sometimes unable to find the globaloptima solution within limit MaxNFFEs Similar to Dim =30 MLBBO outperforms DE on the majority of functionsand overcomes BBO on almost all the functions at everydimension MLBBO is better than or similar to DEBBO onmost of the functions
When compared with aBBOmDE we can see fromTable 5 thatMLBBO is better than or similar to aBBOmDEonmost of the functions too By carefully looking at results wecan recognize that results of some functions of MLBBO areworse than those of aBBOmDE in precision but robustnessof MLBBO is clearly better than aBBOmDE For exampleMLBBO and aBBOmDE find better solution for functionsf8 and f9 when Dim = 10 and 50 aBBOmDE fails to solvefunctions f8 and f9 when Dim increases to 100 and 200 butMLBBO can solve functions f8 and f9 as before This may bedue to modified DE mutation being used in aBBOmDE aslocal search to accelerate the convergence speed butmodifiedDE mutation is used in MLBBO as a formula to supplementwith migration operator and enhance the exploration abilityof MLBBO
ForCEC 2005 test functions F1ndashF14 we can obtain similarresults from Table 6 that MLBBO is better than or similarto DE BBO and DEBBO on most of functions whencomparing MLBBO with these algorithms However resultsof MLBBO outperform aBBOmDE on most of the functionswherever Dim = 10 or Dim = 50
From the comparison we can arrive to a conclusion thatthe total performance of MLBBO is better than the othersespecially for CEC 2005 test functions
Figure 7 shows the NFFEs required to reach VTR withdifferent dimensions for 13 standard benchmark functions(f1ndashf13) and 6 CEC 2005 test functions (F1 F2 F4 F6F7 and F9) From Figure 7 we can see that as dimensionincreases the required NFFEs increases exponentially whichmay be due to an exponentially increase in local minimawith dimension Meanwhile if we compare Figures 7(a) and7(b) we can find that CEC 2005 test functions need moreNFFEs to reach VTR than standard benchmark functionsdo which may be due to that CEC 2005 functions aremore complicated than standard benchmark functions Forexample for function f3 (Schwefel 12) F2 (Shifted Schwefel12) and F4 (Shifted Schwefel 12 with Noise in Fitness) F4F2 and f3 need 1404867 1575967 and 289960 MeanFEsto reach VTR (10minus6) respectively Obviously F4 is morecomplicated than f3 and F2 so F4 needs 289960 MeanFEsto reach VTR which is large than 1404867 and 157596
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
14 Journal of Applied Mathematics
Table 5 Scalability study at Dim = 10 50 100 and 200 a ter Dim lowast 10000 NFFEs for standard functions f1ndashf13
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
dim = 10f1 155119864 minus 88 (653119864 minus 88) 883119864 minus 76 (189119864 minus 75)+ 214119864 minus 02 (204119864 minus 02)+ 000119864 + 00 (000119864 + 00) 149119864 minus 270 (000119864 + 00)f2 282119864 minus 46 (366119864 minus 46) 185119864 minus 36 (175119864 minus 36)+ 102119864 minus 01 (341119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 637119864 minus 45 (280119864 minus 44) 854119864 minus 53 (167119864 minus 52)minus 329119864 + 00 (202119864 + 00)+ 607119864 minus 14 (960119864 minus 14) mdashf4 327119864 minus 31 (418119864 minus 31) 402119864 minus 29 (660119864 minus 29)+ 577119864 minus 01 (257119864 minus 01)minus 158119864 minus 16 (988119864 minus 17) mdashf5 105119864 minus 30 (430119864 minus 30) 000119864 + 00 (000119864 + 00)minus 143119864 + 01 (109119864 + 01)+ 340119864 + 00 (803119864 minus 01) 826119864 + 00 (169119864 + 01)f6 000119864 + 00 (000119864 + 00) 667119864 minus 02 (254119864 minus 01)minus 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdashf7 978119864 minus 04 (739119864 minus 04) 120119864 minus 03 (819119864 minus 04)minus 652119864 minus 04 (736119864 minus 04)sim 111119864 minus 03 (396119864 minus 04) mdashf8 000119864 + 00 (000119864 + 00) 592119864 + 01 (866119864 + 01)+ 599119864 minus 02 (351119864 minus 02)+ 000119864 + 00 (000119864 + 00) 909119864 minus 14 (276119864 minus 13)f9 000119864 + 00 (000119864 + 00) 775119864 + 00 (662119864 + 00)+ 120119864 minus 02 (126119864 minus 02)+ 000119864 + 00 (000119864 + 00) 000119864 + 00 (000119864 + 00)f10 231119864 minus 15 (108119864 minus 15) 266119864 minus 15 (000119864 + 00)minus 660119864 minus 02 (258119864 minus 02)+ 589119864 minus 16 (000119864 + 00) 288119864 minus 15 (852119864 minus 16)f11 394119864 minus 03 (686119864 minus 03) 743119864 minus 02 (545119864 minus 02)+ 948119864 minus 02 (340119864 minus 02)+ 000119864 + 00 (000119864 + 00) 448119864 minus 02 (226119864 minus 02)f12 471119864 minus 32 (167119864 minus 47) 471119864 minus 32 (167119864 minus 47)minus 102119864 minus 03 (134119864 minus 03)+ 471119864 minus 32 (000119864 + 00) 471119864 minus 32 (000119864 + 00)f13 135119864 minus 32 (557119864 minus 48) 285119864 minus 32 (823119864 minus 32)minus 401119864 minus 03 (429119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 50f1 247119864 minus 63 (284119864 minus 63) 442119864 minus 38 (697119864 minus 38)+ 130119864 minus 01 (346119864 minus 02)+ 000119864 + 00 (000119864 + 00) 33119864 minus 154 (14119864 minus 153)f2 854119864 minus 35 (890119864 minus 35) 666119864 minus 19 (604119864 minus 19)+ 554119864 minus 01 (806119864 minus 02)+ 000119864 + 00 (000119864 + 00) mdashf3 123119864 minus 07 (149119864 minus 07) 373119864 minus 06 (623119864 minus 06)+ 787119864 + 02 (176119864 + 02)+ 520119864 minus 02 (248119864 minus 02) mdashf4 208119864 minus 02 (436119864 minus 03) 127119864 + 00 (782119864 minus 01)+ 408119864 + 00 (488119864 minus 01)+ 342119864 minus 03 (228119864 minus 02) mdashf5 225119864 minus 02 (580119864 minus 02) 106119864 + 00 (179119864 + 00)+ 144119864 + 02 (415119864 + 01)+ 443119864 + 01 (139119864 + 01) 950119864 + 00 (269119864 + 01)f6 000119864 + 00 (000119864 + 00) 587119864 + 00 (456119864 + 00)+ 100119864 minus 01 (305119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 507119864 minus 03 (131119864 minus 03) 153119864 minus 02 (501119864 minus 03)+ 291119864 minus 04 (260119864 minus 04)minus 405119864 minus 03 (920119864 minus 04) mdashf8 182119864 minus 11 (000119864 + 00) 141119864 + 04 (344119864 + 02)+ 401119864 minus 01 (138119864 minus 01)+ 237119864 + 00 (167119864 + 01) 239119864 minus 11 (369119864 minus 12)f9 000119864 + 00 (000119864 + 00) 355119864 + 02 (198119864 + 01)+ 701119864 minus 02 (229119864 minus 02)+ 000119864 + 00 (000119864 + 00) 443119864 minus 14 (478119864 minus 14)f10 965119864 minus 15 (355119864 minus 15) 597119864 minus 11 (295119864 minus 10)minus 778119864 minus 02 (134119864 minus 02)+ 592119864 minus 15 (179119864 minus 15) 256119864 minus 14 (606119864 minus 15)f11 131119864 minus 03 (451119864 minus 03) 501119864 minus 03 (635119864 minus 03)+ 155119864 minus 01 (464119864 minus 02)+ 000119864 + 00 (000119864 + 00) 278119864 minus 02 (342119864 minus 02)f12 963119864 minus 33 (786119864 minus 34) 706119864 minus 02 (122119864 minus 01)+ 289119864 minus 04 (203119864 minus 04)+ 942119864 minus 33 (000119864 + 00) 942119864 minus 33 (000119864 + 00)f13 137119864 minus 32 (900119864 minus 34) 105119864 minus 15 (577119864 minus 15)minus 782119864 minus 03 (578119864 minus 03)+ 135119864 minus 32 (000119864 + 00) 135119864 minus 32 (000119864 + 00)
dim = 100f1 653119864 minus 51 (915119864 minus 51) 102119864 minus 34 (106119864 minus 34)+ 289119864 minus 01 (631119864 minus 02)+ 616119864 minus 34 (197119864 minus 33) 118119864 minus 96 (326119864 minus 96)f2 102119864 minus 25 (225119864 minus 25) 820119864 minus 19 (140119864 minus 18)+ 114119864 + 00 (118119864 minus 01)+ 000119864 + 00 (000119864 + 00) mdashf3 137119864 minus 01 (593119864 minus 02) 570119864 + 00 (193119864 + 00)+ 293119864 + 03 (486119864 + 02)+ 310119864 minus 04 (112119864 minus 04) mdashf4 452119864 minus 01 (311119864 minus 01) 299119864 + 01 (374119864 + 00)+ 601119864 + 00 (571119864 minus 01)+ 271119864 + 00 (258119864 + 00) mdashf5 389119864 + 01 (422119864 + 01) 925119864 + 01 (478119864 + 01)+ 322119864 + 02 (674119864 + 01)+ 119119864 + 02 (338119864 + 01) 257119864 + 01 (408119864 + 01)f6 000119864 + 00 (000119864 + 00) 632119864 + 01 (238119864 + 01)+ 667119864 minus 02 (254119864 minus 01)sim 000119864 + 00 (000119864 + 00) mdashf7 180119864 minus 02 (388119864 minus 03) 549119864 minus 02 (135119864 minus 02)+ 192119864 minus 04 (220119864 minus 04)- 687119864 minus 03 (115119864 minus 03) mdashf8 118119864 minus 10 (119119864 minus 11) 320119864 + 04 (523119864 + 02)+ 823119864 minus 01 (207119864 minus 01)+ 711119864 + 00 (284119864 + 01) 107119864 + 02 (113119864 + 02)f9 261119864 minus 13 (266119864 minus 13) 825119864 + 02 (472119864 + 01)+ 137119864 minus 01 (312119864 minus 02)+ 736119864 minus 01 (848119864 minus 01) 159119864 minus 01 (368119864 minus 01)f10 422119864 minus 14 (135119864 minus 14) 225119864 + 00 (492119864 minus 01)+ 826119864 minus 02 (108119864 minus 02)+ 784119864 minus 15 (703119864 minus 16) 425119864 minus 14 (799119864 minus 15)f11 540119864 minus 03 (122119864 minus 02) 369119864 minus 03 (592119864 minus 03)sim 192119864 minus 01 (499119864 minus 02)+ 000119864 + 00 (000119864 + 00) 510119864 minus 02 (889119864 minus 02)f12 189119864 minus 32 (243119864 minus 32) 580119864 minus 01 (946119864 minus 01)+ 270119864 minus 04 (270119864 minus 04)+ 471119864 minus 33 (000119864 + 00) 471119864 minus 33 (000119864 + 00)f13 337119864 minus 32 (221119864 minus 32) 115119864 + 02 (329119864 + 01)+ 108119864 minus 02 (398119864 minus 03)+ 176119864 minus 32 (268119864 minus 32) 155119864 minus 32 (598119864 minus 33)
dim = 200f1 801119864 minus 25 (123119864 minus 24) 865119864 minus 27 (272119864 minus 26)minus 798119864 minus 01 (133119864 minus 01)+ 307119864 minus 32 (248119864 minus 32) 570119864 minus 50 (734119864 minus 50)f2 512119864 minus 05 (108119864 minus 04) 358119864 minus 14 (743119864 minus 14)minus 255119864 + 00 (171119864 minus 01)+ 583119864 minus 17 (811119864 minus 17) mdashf3 639119864 + 01 (105119864 + 01) 828119864 + 02 (197119864 + 02)+ 904119864 + 03 (712119864 + 02)+ 222119864 + 05 (590119864 + 04) mdashf4 713119864 + 00 (163119864 + 00) 427119864 + 01 (311119864 + 00)+ 884119864 + 00 (649119864 minus 01)+ 159119864 + 01 (343119864 + 00) mdashf5 309119864 + 02 (635119864 + 01) 323119864 + 02 (813119864 + 01)sim 629119864 + 02 (626119864 + 01)+ 295119864 + 02 (448119864 + 01) 216119864 + 02 (998119864 + 01)f6 000119864 + 00 (000119864 + 00) 936119864 + 02 (450119864 + 02)+ 000119864 + 00 (000119864 + 00)sim 000119864 + 00 (000119864 + 00) mdash
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 15
Table 5 Continued
MLBBO DE BBO DEBBO aBBOmDEMean (std) Mean (std) Mean (std) Mean (std) Mean (std)
f7 807119864 minus 02 (124119864 minus 02) 211119864 minus 01 (540119864 minus 02)+ 146119864 minus 04 (106119864 minus 04)minus 156119864 minus 02 (282119864 minus 03) mdashf8 114119864 minus 09 (155119864 minus 09) 696119864 + 04 (612119864 + 02)+ 213119864 + 00 (330119864 minus 01)+ 201119864 + 02 (168119864 + 02) 308119864 + 02 (222119864 + 02)
f9 975119864 minus 12 (166119864 minus 11) 338119864 + 02 (206119864 + 02)+ 325119864 minus 01 (398119864 minus 02)+ 176119864 + 01 (289119864 + 00) 119119864 + 00 (752119864 minus 01)
f10 530119864 minus 13 (111119864 minus 12) 600119864 + 00 (109119864 + 00)+ 992119864 minus 02 (967119864 minus 03)+ 109119864 minus 14 (112119864 minus 15) 109119864 minus 13 (185119864 minus 14)
f11 529119864 minus 03 (173119864 minus 02) 283119864 minus 02 (757119864 minus 02)sim 288119864 minus 01 (572119864 minus 02)+ 111119864 minus 16 (000119864 + 00) 554119864 minus 02 (110119864 minus 01)
f12 190119864 minus 15 (583119864 minus 15) 236119864 + 00 (235119864 + 00)+ 309119864 minus 04 (161119864 minus 04)+ 236119864 minus 33 (000119864 + 00) 236119864 minus 33 (000119864 + 00)
f13 225119864 minus 25 (256119864 minus 25) 401119864 + 02 (511119864 + 01)+ 317119864 minus 02 (713119864 minus 03)+ 836119864 minus 32 (144119864 minus 31) 135119864 minus 32 (000119864 + 00)
Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
MLBBOOBBOPBBO
MOBBORCBBO
0 05 1 15 2 25 3 3510
13
1014
1015
1016
Best
fitne
ss
F11
times105
Number of fitness function evaluations
05 1 15 2 25 3 3510
minus2
100
102
F7
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus30
10minus20
10minus10
100
1010 F1
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Rastrigin
times105
Number of fitness function evaluations
Best
fitne
ss
0 5 10 15
10minus15
10minus10
10minus5
100
Ackley
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 2510
minus4
10minus2
100
102
104 Griewank
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2 25 3 3510
minus15
10minus10
10minus5
100
105
Schwefel
times105
Number of fitness function evaluationsBe
st fit
ness
0 1 2 3 4 5 610
minus8
10minus6
10minus4
10minus2
100
102
Schwefel4
times105
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 2times10
5
10minus40
10minus30
10minus20
10minus10
100
1010
Number of fitness function evaluations
Sphere
Best
fitne
ss
Figure 4 Convergence curves of mean fitness values of MLBBO OBBO PBBO MOBBO and RCBBO for functions f1 f4 f8 f9 f6 f7 f8 f9f10 F1 F7 and F11
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
16 Journal of Applied Mathematics
0 1 2 3 4 5 6times10
5
10minus20
10minus10
100
1010
Number of fitness function evaluations
Best
fitne
ss
Schwefel2
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
MLBBOOXDEDEBBO
0 1 2 3 4 5 6times10
5
Number of fitness function evaluations
10minus30
10minus20
10minus10
100
1010
Rosenbrock
Best
fitne
ss
0 2 4 6 8 1010
minus2
100
102
104
106
Step
times104
Number of fitness function evaluations
Best
fitne
ss
0 05 1 15 210
minus40
10minus20
100
1020
Penalized1
times105
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3 410
4
106
108
1010
F3
Best
fitne
ss
times105
Number of fitness function evaluations0 1 2 3 4
102
103
104
105
F5
times105
Number of fitness function evaluations
Best
fitne
ss
10minus20
10minus10
100
1010
F9
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
101
102
103
F10
0 1 2 3 4times10
5
Number of fitness function evaluations
Best
fitne
ss
0 1 2 3
10111
10113
10115
F14
times105
Number of fitness function evaluations
Best
fitne
ss
Figure 5 Convergence curves of mean fitness values of MLBBO OXDE and DEBBO for functions f3 f5 f6 f12 F3 F5 F9 F10 and F14
47 Analysis of the Performance of Operators in MLBBO Asthe analysis above the proposed variant of BBO algorithmhas achieved excellent performance In this section weanalyze the performance of modified migration operatorand local search mechanism in MLBBO In order to ana-lyze the performance fairly and systematically we comparethe performance in four cases MLBBO with both mod-ified migration operator and local search leaning mecha-nism MLBBO with the modified migration operator andwithout local search leaning mechanism MLBBO withoutmodified migration operator and with local search leaningmechanism and MLBBO without both modified migra-tion operator and local search leaning mechanism (it isactually BBO) Four algorithms are denoted as MLBBOMLBBO2 MLBBO3 and MLBBO4 The experimental testsare conducted with 30 independent runs The parametersettings are set in Section 41 The mean standard deviation
of fitness values of experimental test are summarized inTable 7 The numbers of successful runs are also recordedin Table 7 The results between MLBBO and MLBBO2MLBBO and MLBBO3 and MLBBO and MLBBO4 arecompared using two tails 119905-test Convergence curves ofmean fitness values under 30 independent runs are listed inFigure 8
From Table 7 we can find that MLBBO is significantlybetter thanMLBBO2 on 15 functions out of 27 tests functionswhereas MLBBO2 outperforms MLBBO on 5 functions Forthe rest 7 functions there are no significant differences basedon two tails 119905-test From this we know that local searchmechanism has played an important role in improving thesearch ability especially in improving the solution precisionwhich can be proved through careful looking at fitnessvalues For example for function f1ndashf10 both algorithmsMLBBO and MLBBO2 achieve the optima but the fitness
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 17
0 2 4 6 8 10times10
5
10minus50
100
1050
10100
Schwefel3
Best
fitne
ss
(1) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus10
10minus5
100
105
Schwefel
times105
Best
fitne
ss
(2) NFFEs (Dim = 100)
0 5 1010
minus40
10minus20
100
1020
Penalized1
times105
Best
fitne
ss
(3) NFFEs (Dim = 100)
0 05 1 15 210
0
102
104
106
Schwefel2
times106
Best
fitne
ss
(4) NFFEs (Dim = 200)0 05 1 15 2
10minus20
10minus10
100
1010
Ackley
times106
Best
fitne
ss
(5) NFFEs (Dim = 200)0 05 1 15 2
10minus40
10minus20
100
1020
Penalized2
times106
Best
fitne
ss
(6) NFFEs (Dim = 200)
0 2 4 6 8 1010
minus40
10minus20
100
1020 F1
times104
Best
fitne
ss
(7) NFFEs (Dim = 10)0 2 4 6 8 10
10minus15
10minus10
10minus5
100
105 F5
times104
Best
fitne
ss
(8) NFFEs (Dim = 10)0 2 4 6 8 10
10131
10132
F8
times104
Best
fitne
ss(9) NFFEs (Dim = 10)
0 2 4 6 8 1010
minus1
100
101
102
103 F13
times104
Best
fitne
ss
(10) NFFEs (Dim = 10)0 1 2 3 4 5
105
1010 F3
times105
Best
fitne
ss
(11) NFFEs (Dim = 50)
0 1 2 3 4 510
1
102
103
104 F10
times105
Best
fitne
ss
(12) NFFEs (Dim = 50)
0 1 2 3 4 5
1016
1017
1018
1019
F11
times105
Best
fitne
ss
(13) NFFEs (Dim = 50)0 1 2 3 4 5
103
104
105
106
107
F12
times105
Best
fitne
ss
(14) NFFEs (Dim = 50)0 2 4 6 8 10
10minus5
100
105
1010
F2
times105
Best
fitne
ss
(15) NFFEs (Dim = 100)
0 2 4 6 8 10106
108
1010
1012
F3
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(16) NFFEs (Dim = 100)
0 2 4 6 8 10104
105
106
107
F4
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(17) NFFEs (Dim = 100)
0 2 4 6 8 1010
minus15
10minus10
10minus5
100
105 F9
MLBBODE
BBODEBBO
Best
fitne
ss
times105
(18) NFFEs (Dim = 100)
Figure 6 Mean fitness curves to compare the scalability of MLBBO DE BBO and DEBBO for selected functions (1) f2 (Dim = 100) (2)f8 (Dim = 100) (3) f12 (Dim = 100) (4) f3 (Dim = 200) (5) f10 (Dim = 200) (6) f13 (Dim = 200) (7) F1 (Dim = 10) (8) F5 (Dim = 10) (9) F8(Dim = 10) (10) F13 (Dim = 10) (11) F3 (Dim = 50) (12) F10 (Dim = 50) (13) F11 (Dim = 50) (14) F12 (Dim = 50) (15) F2 (Dim = 100) (16)F3 (Dim = 100) (17) F4 (Dim = 100) and(18) F9 (Dim = 100)
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
18 Journal of Applied Mathematics
0 50 100 150 2000
2
4
6
8
10
12
14
16
18times10
5
NFF
Es
f1f2f3f4f5f6f7
f8f9f10f11f12f13
(a) Dimension
0 20 40 60 80 1000
1
2
3
4
5
6
7
8
NFF
Es
F1F2F4
F6F7F9
times105
(b) Dimension
Figure 7 NFFEs versus dimensions for (a) standard functions and (b) CEC 2005 functions
0 5 10 15times10
4
10minus40
10minus20
100
1020
Number of fitness function evaluations
Best
fitne
ss
Sphere
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
MLBBOMLBBO2
MLBBO3MLBBO4
10minus2
100
102
104
106
Step
0 1 2 3times10
4
Number of fitness function evaluations
Best
fitne
ss
10minus15
10minus10
10minus5
100
105
Rastrigin
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
1014
1015
1016
F11
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
104
106
108
1010 F3
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
10minus40
10minus20
100
1020 F1
0 1 2 3times10
5
Number of fitness function evaluations
Best
fitne
ss
Figure 8 Mean fitness curves to analyze operators of MLBBO for selected functions f1 f6 f9 F1 F3 and F11
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 19
Table 6 Scalability study at Dim = 10 50 and 100 after Dim lowast 10000 NFFEs for CEC 2005 test functions F1ndashF14
MLBBO DE BBO DEBBO aBBOmDEMean(std) Mean(std) Mean(std) Mean(std) Mean(std)
dim = 10F1 000119864 + 00 (000119864 + 00) 471119864 minus 29 (806119864 minus 29)
+186119864 minus 02 (976119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim000119864 + 00 (000119864 + 00)
F2 711119864 minus 29 (742119864 minus 29) 631119864 minus 29 (757119864 minus 29)sim373119864 + 01 (375119864 + 01)
+850119864 minus 25 (282119864 minus 24)
sim365119864 minus 23(258119864 minus 22)
F3 840119864 + 01 (419119864 + 02) 805119864 minus 25(916119864 minus 25)sim 237119864 + 06 (301119864 + 06)+ 186119864 + 04 (353119864 + 04)+ 102119864 + 05 (977119864 + 04)F4 728119864 minus 29 (957119864 minus 29) 391119864 minus 29 (705119864 minus 29)
sim301119864 + 02 (331119864 + 02)
+227119864 minus 21 (378119864 minus 21)
+196119864 minus 02 (619119864 minus 02)
F5 300119864 minus 12 (510119864 minus 12) 197119864 minus 12 (241119864 minus 12)sim168119864 + 03 (109119864 + 03)
+288119864 minus 05 (148119864 minus 04)
+408119864 + 02 (693119864 + 02)
F6 498119864 minus 26 (194119864 minus 25) 132119864 minus 26 (178119864 minus 26)sim126119864 + 02 (135119864 + 02)
+410119864 + 00 (220119864 + 00)
+140119864 + 02 (822119864 + 02)
F7 587119864 minus 02 (435119864 minus 02) 135119864 minus 01 (144119864 minus 01)+133119864 + 00 (404119864 minus 01)
+386119864 minus 02 (236119864 minus 02)
minus115119864 + 00 (108119864 + 00)
F8 204119864 + 01 (780119864 minus 02) 203119864 + 01 (732119864 minus 02)sim204119864 + 01 (125119864 minus 01)
sim204119864 + 01 (513119864 minus 02)
sim203119864 + 01 (835119864 minus 02)
F9 000119864 + 00 (000119864 + 00) 839119864 + 00 (694119864 + 00)+888119864 minus 03 (451119864 minus 03)
+000119864 + 00 (000119864 + 00)
sim312119864 minus 09 (221119864 minus 08)
F10 616119864 + 00 (297119864 + 00) 229119864 + 01 (574119864 + 00)+182119864 + 01 (869119864 + 00)
+598119864 + 00 (242119864 + 00)
sim157119864 + 01 (604119864 + 00)
F11 172119864 + 00 (159119864 + 00) 248119864 + 00 (386119864 + 00)sim693119864 + 00 (143119864 + 00)
+828119864 minus 01 (139119864 + 00)
minus mdashF12 347119864 + 02 (608119864 + 02) 255119864 + 02 (530119864 + 02)
sim723119864 + 02 (748119864 + 02)
+104119864 + 02 (340119864 + 02)
sim mdashF13 556119864 minus 01 (744119864 minus 02) 181119864 + 00 (608119864 minus 01)
+362119864 minus 01 (135119864 minus 01)
minus536119864 minus 01 (482119864 minus 02)
sim mdashF14 242119864 + 00 (492119864 minus 01) 254119864 + 00 (302119864 minus 01)
sim359119864 + 00 (382119864 minus 01)
+309119864 + 00 (176119864 minus 01)
+ mdashdim = 50
F1 825119864 minus 29 (941119864 minus 29) 177119864 minus 27 (583119864 minus 28)+118119864 minus 01 (322119864 minus 02)
+000119864 + 00 (000119864 + 00)
minus000119864 + 00 (000119864 + 00)
F2 911119864 minus 07 (991119864 minus 07) 106119864 minus 04 (175119864 minus 04)+125119864 + 04 (369119864 + 03)
+104119864 + 03 (304119864 + 02)
+262119864 minus 02 (717119864 minus 02)
F3 186119864 + 05 (684119864 + 04) 549119864 + 05 (201119864 + 05)+193119864 + 07 (590119864 + 06)
+720119864 + 06 (238119864 + 06)
+133119864 + 06 (489119864 + 05)
F4 173119864 + 01 (158119864 + 01) 319119864 + 02 (252119864 + 02)+436119864 + 04 (119119864 + 04)
+726119864 + 03 (217119864 + 03)
+129119864 + 04 (715119864 + 03)
F5 418119864 + 03 (624119864 + 02) 266119864 + 03 (509119864 + 02)minus116119864 + 04 (163119864 + 03)
+288119864 + 03 (422119864 + 02)
minus140119864 + 04 (267119864 + 03)
F6 288119864 + 00 (148119864 + 01) 116119864 + 00 (183119864 + 00)sim305119864 + 02 (839119864 + 01)
+536119864 + 01 (214119864 + 01)
+406119864 + 01 (566119864 + 01)
F7 141119864 minus 02 (145119864 minus 02) 845119864 minus 03 (106119864 minus 02)sim123119864 + 00 (907119864 minus 02)
+279119864 minus 03 (655119864 minus 03)
minus115119864 minus 02 (155119864 minus 02)
F8 211119864 + 01 (323119864 minus 02) 211119864 + 01 (337119864 minus 02)sim210119864 + 01 (821119864 minus 02)
minus211119864 + 01 (421119864 minus 02)
sim211119864 + 01 (539119864 minus 02)
F9 474119864 minus 16 (114119864 minus 15) 383119864 + 02 (357119864 + 01)+667119864 minus 02 (201119864 minus 02)
+332119864 minus 02 (182119864 minus 01)
sim222119864 minus 08 (157119864 minus 07)
F10 866119864 + 01 (222119864 + 01) 430119864 + 02 (336119864 + 01)+998119864 + 01 (282119864 + 01)
+163119864 + 02 (135119864 + 01)
+148119864 + 02 (415119864 + 01)
F11 360119864 + 01 (840119864 + 00) 728119864 + 01 (186119864 + 00)+585119864 + 01 (462119864 + 00)
+592119864 + 01 (144119864 + 00)
+ mdashF12 677119864 + 03 (532119864 + 03) 737119864 + 03 (805119864 + 03)
sim515119864 + 04 (261119864 + 04)
+105119864 + 04 (938119864 + 03)
sim mdashF13 556119864 + 00 (276119864 minus 01) 325119864 + 01 (206119864 + 00)
+186119864 + 00 (258119864 minus 01)
minus523119864 + 00 (288119864 minus 01)
minus mdashF14 223119864 + 01 (353119864 minus 01) 230119864 + 01 (180119864 minus 01)
+227119864 + 01 (376119864 minus 01)
+226119864 + 01 (162119864 minus 01)
+ mdashdim = 100
F1 798119864 minus 28 (115119864 minus 27) 680119864 minus 27 (204119864 minus 27)+299119864 minus 01 (626119864 minus 02)
+168119864 minus 29 (519119864 minus 29)
minus mdashF2 561119864 minus 01 (238119864 minus 01) 105119864 + 02 (454119864 + 01)
+398119864 + 04 (734119864 + 03)
+242119864 + 04 (508119864 + 03)
+ mdashF3 202119864 + 06 (535119864 + 05) 206119864 + 06 (713119864 + 05)
sim452119864 + 07 (983119864 + 06)
+142119864 + 07 (415119864 + 06)
+ mdashF4 163119864 + 04 (827119864 + 03) 431119864 + 04 (113119864 + 04)
+148119864 + 05 (252119864 + 04)
+847119864 + 04 (116119864 + 04)
+ mdashF5 132119864 + 04 (140119864 + 03) 811119864 + 03 (129119864 + 03)
minus301119864 + 04 (332119864 + 03)
+608119864 + 03 (628119864 + 02)
minus mdashF6 398119864 + 01 (359119864 + 01) 124119864 + 02 (505119864 + 01)
+474119864 + 02 (532119864 + 01)
+115119864 + 02 (281119864 + 01)
+ mdashF7 369119864 minus 03 (649119864 minus 03) 402119864 minus 03 (619119864 minus 03)
sim121119864 + 00 (638119864 minus 02)
+148119864 minus 03 (397119864 minus 03)
sim mdashF8 213119864 + 01 (216119864 minus 02) 213119864 + 01 (232119864 minus 02)
sim211119864 + 01 (563119864 minus 02)
minus213119864 + 01 (205119864 minus 02)
sim mdashF9 103119864 minus 14 (470119864 minus 15) 780119864 + 02 (306119864 + 02)
+134119864 minus 01 (310119864 minus 02)
+156119864 + 00 (110119864 + 00)
+ mdashF10 296119864 + 02 (514119864 + 01) 110119864 + 03 (450119864 + 01)
+274119864 + 02 (377119864 + 01)
sim397119864 + 02 (250119864 + 01)
+ mdashNote ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
values obtained by MLBBO are better than those obtainedby MLBBO2 by several orders of magnitudes on thesefunctions When compared with MLBBO3 we can see thatMLBBO outperforms MLBBO3 on 19 functions out of 27test functions while MLBBO is outperformed by MLBBO3
on only 2 functions From this we know that the modifiedmigration operator has played a key role in optimizing theprocess and is better than originalmigration operator Similarconclusion can also be obtained if we compare MLBBO2 andMLBBO4 MLBBO3 and MLBBO4
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
20 Journal of Applied Mathematics
Table 7 Analysis of the performance of operators in MLBBO
MLBBO MLBBO2 MLBBO3 MLBBO4Mean (std) SR Mean (std) SR Mean (std) SR Mean (std) SR
f1 203119864 minus 31 (177119864 minus 31) 30 295119864 minus 18 (105119864 minus 18)+ 30 453119864 minus 05 (265119864 minus 05)
+ 0 217119864 minus 04 (792119864 minus 05)+ 0
f2 160119864 minus 21 (848119864 minus 22) 30 440119864 minus 13 (113119864 minus 13)+ 30 752119864 minus 03 (274119864 minus 03)
+ 0 364119864 minus 02 (700119864 minus 03)+ 0
f3 147119864 minus 20 (210119864 minus 20) 30 294119864 minus 12 (194119864 minus 12)+ 30 188119864 + 00 (610119864 minus 01)
+ 0 198119864 + 00 (563119864 minus 01)+ 0
f4 504119864 minus 08 (988119864 minus 08) 30 304119864 minus 09 (122119864 minus 09)minus 30 196119864 minus 02 (467119864 minus 03)
+ 0 232119864 minus 02 (647119864 minus 03)+ 0
f5 102119864 minus 20 (371119864 minus 20) 30 154119864 minus 14 (106119864 minus 14)+ 30 479119864 + 01 (328119864 + 01)
+ 0 364119864 + 01 (358119864 + 01)+ 0
f6 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 000119864 + 00 (000119864 + 00)
sim 30 000119864 + 00 (000119864 + 00)sim 30
f7 263119864 minus 03 (124119864 minus 03) 30 400119864 minus 03 (768119864 minus 04)+ 30 325119864 minus 03 (164119864 minus 03)
sim 30 128119864 minus 02 (507119864 minus 03)+ 11
f8 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 163119864 minus 06 (125119864 minus 06)
+ 6 792119864 minus 06 (284119864 minus 06)+ 0
f9 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 793119864 minus 04 (464119864 minus 04)
+ 0 360119864 minus 03 (146119864 minus 03)+ 0
f10 598119864 minus 15 (901119864 minus 16) 30 448119864 minus 10 (116119864 minus 10)+ 30 432119864 minus 03 (124119864 minus 03)
+ 0 972119864 minus 03 (150119864 minus 03)+ 0
f11 370119864 minus 03 (602119864 minus 03) 19 000119864 + 00 (000119864 + 00)minus 30 107119864 minus 01 (687119864 minus 02)
+ 1 816119864 minus 02 (571119864 minus 02)+ 0
f12 569119864 minus 27 (307119864 minus 26) 30 441119864 minus 20 (249119864 minus 20)+ 30 166119864 minus 06 (543119864 minus 06)
sim 26 768119864 minus 06 (128119864 minus 05)+ 1
f13 579119864 minus 32 (553119864 minus 32) 30 360119864 minus 19 (165119864 minus 19)+ 30 333119864 minus 05 (116119864 minus 04)
sim 0 570119864 minus 05 (507119864 minus 05)+ 0
F1 000119864 + 00 (000119864 + 00) 30 000119864 + 00 (000119864 + 00)sim 30 948119864 minus 06 (828119864 minus 06)
+ 2 466119864 minus 05 (191119864 minus 05)+ 0
F2 237119864 minus 12 (246119864 minus 12) 30 279119864 minus 06 (244119864 minus 06)+ 4 410119864 + 01 (107119864 + 01)
+ 0 239119864 + 01 (101119864 + 01)+ 0
F3 948119864 + 04 (514119864 + 04) 0 186119864 + 05 (984119864 + 04)+ 0 186119864 + 06 (575119864 + 05)
+ 0 951119864 + 06 (757119864 + 06)+ 0
F4 121119864 minus 04 (276119864 minus 04) 4 657119864 minus 03 (898119864 minus 03)+ 0 112119864 + 04 (405119864 + 03)
+ 0 489119864 + 03 (177119864 + 03)+ 0
F5 818119864 + 02 (325119864 + 02) 0 127119864 + 02 (659119864 + 01)minus 0 609119864 + 03 (112119864 + 03)
+ 0 447119864 + 03 (700119864 + 02)+ 0
F6 193119864 minus 09 (311119864 minus 09) 30 782119864 minus 04 (405119864 minus 03)sim 29 926119864 + 02 (186119864 + 03)
+ 0 209119864 + 03 (448119864 + 03)+ 0
F7 189119864 minus 02 (173119864 minus 02) 13 156119864 minus 03 (441119864 minus 03)minus 29 167119864 minus 01 (206119864 minus 01)
+ 0 776119864 minus 01 (139119864 + 00)+ 0
F8 210119864 + 01 (538119864 minus 02) 0 209119864 + 01 (606119864 minus 02)sim 0 209119864 + 01 (410119864 minus 02)
sim 0 210119864 + 01 (380119864 minus 02)sim 0
F9 592119864 minus 17 (324119864 minus 16) 30 000119864 + 00 (000119864 + 00)sim 30 106119864 minus 03 (671119864 minus 04)
+ 30 344119864 minus 03 (145119864 minus 03)+ 30
F10 376119864 + 01 (140119864 + 01) 0 974119864 + 01 (654119864 + 00)+ 0 379119864 + 01 (105119864 + 01)
sim 0 122119864 + 02 (819119864 + 00)+ 0
F11 210119864 + 01 (737119864 + 00) 0 321119864 + 01 (105119864 + 00)+ 0 263119864 + 01 (496119864 + 00)
+ 0 334119864 + 01 (107119864 + 00)+ 0
F12 207119864 + 03 (271119864 + 03) 0 136119864 + 04 (197119864 + 04)+ 0 126119864 + 04 (929119864 + 03)
+ 0 805119864 + 03 (703119864 + 03)+ 0
F13 308119864 + 00 (160119864 minus 01) 0 277119864 + 00 (138119864 minus 01)minus 0 104119864 + 00 (164119864 minus 01)
minus 0 982119864 minus 01 (174119864 minus 01)minus 0
F14 128119864 + 01 (238119864 minus 01) 0 130119864 + 01 (162119864 minus 01)+ 0 125119864 + 01 (265119864 minus 01)
minus 0 130119864 + 01 (186119864 minus 01)+ 0
Better 15 19 24Similar 7 6 2Worse 5 2 1Note ldquo+rdquo ldquominusrdquo and ldquosimrdquo indicate that MLBBO is significantly better worse and indifferent respectively compared with other algorithms
Furthermore if we compare four algorithms togetherthe results of MLBBO4 are worse than those of MLBBOespecially on multimodal functions which means thatthe MLBBO has more powerful exploration ability thanMLBBO4
5 Conclusions and Future Work
In this paper we have proposed a variant of modified BBOalgorithm called MLBBO for solving global optimizationproblems For origin BBO algorithm the exploration abilityis a barrier of performance of BBO We suggest a modifiedmigration operator by combining the formula in migrationoperator with a new one which can absorbmore informationfrom other habitats and then enhance the exploration abilityMoreover a local search mechanism is designed and used
in modified BBO to supplement with modified migrationoperator The proposed algorithm is compared with otherimproved BBO algorithms and evolutionary algorithmswhich show that the proposed algorithm is superior to or atleast highly competitive with the others
During the analysis we know that MLBBO possessesbetter performance and we will investigate the performanceof MLBBO in large-scale functions in the future work Wewill also investigate the built-in relationship between thepopulation diversity and exploration ability further
6 Appendix
See Table 8
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 21
Table8Be
nchm
arkfunctio
nsused
inou
rexp
erim
entaltests
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f11198911(119909)=
119899
sum 119894=1
1199092 119894
Sphere
n[minus100100]
0
f21198912(119909)=
119899
sum 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816+
119899
prod 119894=1
1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816Schw
efelrsquosP
roblem
222
n[1010]
0
f31198913(119909)=
119899
sum 119894=1
(
119894
sum 119895=1
119909119895)2
Schw
efelrsquosP
roblem
12n
[minus100100]
0
f41198914(119909)=max1198941003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 10038161le119894le119899
Schw
efelrsquosP
roblem
221
n[minus100100]
0
f51198915(119909)=
119899minus1
sum 119894=1
[100(119909119894+1minus1199092 119894)2+(119909119894minus1)2]
Rosenb
rock
functio
nn
[minus3030]
0
f61198916(119909)=
119899
sum 119894=1
(lfloor119909119894+05rfloor)2
Step
functio
nn
[minus100100]
0
f71198917(119909)=
119899
sum 119894=1
1198941199094 119894+rand
om[01)
Quarticfunctio
nn
[minus12
812
8]0
f81198918(119909)=minus
119899
sum 119894=1
119909119894sin(radic1003816 1003816 1003816 1003816119909119894
1003816 1003816 1003816 1003816)Schw
efelfunctio
nn
[minus512512]
minus125695
f91198919(119909)=
119899
sum 119894=1
[1199092 119894minus10cos(2120587119909119894)+10]
Rastrig
infunctio
nn
[minus512512]
0
f10
11989110(119909)=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199092 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119909119894)+20+119890
Ackley
functio
nn
[minus3232]
0
f11
11989111(119909)=
1
4000
119899
sum 119894=1
1199092 119894minus
119899
prod 119894=1
cos (119909119894
radic119894
)+1
Grie
wankfunctio
nn
[minus60
060
0]0
f12
11989112(119909)=120587 11989910sin2(120587119910119894)+
119899minus1
sum 119894=1
(119910119894minus1)2[1+10sin2(120587119910119894+1)]
+(119910119899minus1)2+
119899
sum 119894=1
119906(119909119894101004)
119910119894=1+(119909119894minus1)
4
119906(119909119894119886119896119898)=
119896(119909119894minus119886)2
0 119896(minus119909119894minus119886)2
119909119894gt119886
minus119886le119909119894le119886
119909119894ltminus119886
Penalized
functio
n1
n[minus5050]
0
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
22 Journal of Applied Mathematics
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
f13
11989113(119909)=01sin2(31205871199091)+
119899minus1
sum 119894=1
(119909119894minus1)2[1+sin2(3120587119909119894+1)]
+(119909119899minus1)2[1+sin2(2120587119909119899)]+
119899
sum 119894=1
119906(11990911989451004)
Penalized
functio
n2
n[minus5050]
0
F11198651=
119899
sum 119894=1
1199112 119894+119891
bias1119885=119883minus119874
Shifted
sphere
functio
nn
[minus100100]
minus450
F21198652=
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)2+119891
bias2119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12n
[minus100100]
minus450
F31198653=
119899
sum 119894=1
(106)(119894minus1)(119863minus1)1199112 119894+119891
bias3119885=119883minus119874
Shifted
rotatedhigh
cond
ition
edellip
ticfunctio
nn
[minus100100]
minus450
F41198654=(
119899
sum 119894=1
(
119894
sum 119895=1
119911119895)
2
)lowast(1+04|119873(01)|)+119891
bias4119885=119883minus119874
Shifted
Schw
efelrsquosP
roblem
12with
noise
infitness
n[minus100100]
minus450
F51198655=max1003816 1003816 1003816 1003816119860119894119909minus119861119894
1003816 1003816 1003816 1003816+119891
bias5
Schw
efelrsquosP
roblem
26with
glob
alop
timum
onbo
unds
n[minus3030]
minus310
F61198656=
119899minus1
sum 119894=1
[100(119911119894+1minus1199112 119894)2+(119911119894minus1)2]+119891
bias6119885=119883minus119874+1
Shifted
rosenb
rockrsquos
functio
nn
[minus100100]
390
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Journal of Applied Mathematics 23
Table8Con
tinued
Functio
nsMathematicalform
ula
Nam
eDim
ensio
nSearch
space
Optim
a
F71198657=
1
4000
119899
sum 119894=1
1199112 119894minus
119899
prod 119894=1
cos(119911119894
radic119894
)+1+119891
bias7119885=(119883minus119874)lowast119872
Shifted
rotatedGrie
wankrsquos
functio
nwith
outb
ound
sn
Outsid
eof
initializa-
tionrange
[060
0]
minus180
F81198658=minus20exp(minus02radic1 119899
119899
sum 119894=1
1199112 119894)minusexp(1 119899
119899
sum 119894=1
cos2120587119911119894)+20+119890+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedAc
kleyrsquos
functio
nwith
glob
alop
timum
onbo
unds
n[minus3232]
minus140
F91198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=119883minus119874
Shifted
Rastrig
inrsquosfunctio
nn
[minus512512]
minus330
F10
1198659=
119899
sum 119894=1
[1199112 119894minus10cos(2120587119911119894)+10]+119891
bias8119885=(119883minus119874)lowast119872
Shifted
rotatedRa
strig
inrsquos
functio
nn
[minus512512]
minus330
F11
11986511=
119899
sum 119894=1
(
119896maxsum 119896=0
[119886119896cos(2120587119887119896(119911119894+05))])minus119899
119896maxsum 119896=0
[119886119896cos(2120587119887119896sdot05)]+119891
bias11
119886=05119887=3119896max=20119885=(119883minus119874)lowast119872
Shifted
rotatedweierstrass
Functio
nn
[minus60
060
0]90
F12
11986512=
119899
sum 119894=1
(119860119894minus119861119894(119909))
2
+119891
bias12
119860119894=
119899
sum 119894=1
(119886119894119895sin120572119895+119887119894119895cos120572119895)119861119894=
119899
sum 119894=1
(119886119894119895sin119909119895+119887119894119895cos119909119895)
Schw
efelrsquosP
roblem
213
n[minus100100]
minus46
0
F13
11986513=11989111(1198915(11991111199112))+11989111(1198915(11991121199113))+sdotsdotsdot+11989111(1198915(119911119899minus1119911119899))+11989111(1198915(1199111198991199111))+119891
bias13
119885=119883minus119874+1
Expand
edextend
edGrie
wankrsquos
plus
Rosenb
rockrsquosfunctio
n(F8F
2)
n[minus5050]
minus130
F14
119865(119909119910)=05+
sin2(radic1199092+1199102)minus05
(1+0001(1199092+1199102))2
11986514=119865(11991111199112)+119865(11991121199113)+sdotsdotsdot+119865(119911119899minus1119911119899)+119865(1199111198991199111)+119891
bias14119885=(119883minus119874)lowast119872
Shifted
rotatedexpand
edScafferrsquosF6
n[minus100100]
minus300
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
24 Journal of Applied Mathematics
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
Acknowledgments
The author would like to express thanks to the area editorand anonymous reviewers for their valuable comments andsuggestions for this paper This work is proudly supportedin part by the National Natural Science Foundation ofChina (No 11101101 No 11226219 No 61373174 and No11361019) Natural Science Foundation of Guangxi (No2013GXNSFBA019008 and No 2013GXNSFAA019003) Sci-ence and Technology Plan Project of Science and TechnologyDepartment of Hunan (No 2013FJ3083) and Project sup-ported by Program to Sponsor Teams for Innovation in theConstruction of Talent Highlands in Guangxi Institutions ofHigher Learning ([2011] 47)
References
[1] E L Lawler and D E Wood ldquoBranch-and-bound methods asurveyrdquo Operations Research vol 14 pp 699ndash719 1966
[2] N Andrei ldquoAccelerated scaled memoryless BFGS precondi-tioned conjugate gradient algorithm for unconstrained opti-mizationrdquo European Journal of Operational Research vol 204no 3 pp 410ndash420 2010
[3] I Ciornei and E Kyriakides ldquoHybrid ant colony-geneticalgorithm (GAAPI) for global continuous optimizationrdquo IEEETransactions on Systems Man and Cybernetics B vol 42 no 1pp 234ndash245 2012
[4] G Chen C P Low and Z Yang ldquoPreserving and exploitinggenetic diversity in evolutionary programming algorithmsrdquoIEEE Transactions on Evolutionary Computation vol 13 no 3pp 661ndash673 2009
[5] C Li S Yang and T T Nguyen ldquoA self-learning particle swarmoptimizer for global optimization problemsrdquo IEEE Transactionson Systems Man and Cybernetics B vol 42 no 3 pp 627ndash6462012
[6] S L Ho S Yang Y Bai and J Huang ldquoAn ant colony algorithmfor both robust and global optimizations of inverse problemsrdquoIEEE Transactions on Magnetics vol 49 no 5 pp 2077ndash20882013
[7] R Storn and K Price ldquoDifferential evolutionmdasha simple andefficient heuristic for global optimization over continuousspacesrdquo Journal of Global Optimization vol 11 no 4 pp 341ndash359 1997
[8] D Simon ldquoBiogeography-based optimizationrdquo IEEE Transac-tions on Evolutionary Computation vol 12 no 6 pp 702ndash7132008
[9] W Gong Z Cai C X Ling and H Li ldquoA real-codedbiogeography-based optimization with mutationrdquo AppliedMathematics and Computation vol 216 no 9 pp 2749ndash27582010
[10] Z Cai W Gong and C X Ling ldquoResearch on a novelbiogeography-based optimization algorithm based on evolu-tionary programmingrdquo System EngineeringTheory and Practicevol 30 no 6 pp 1106ndash1112 2010
[11] I Boussaıd A Chatterjee P Siarry and M Ahmed-NacerldquoTwo-stage update biogeography-based optimization using dif-ferential evolution algorithm (DBBO)rdquo Computers and Opera-tions Research vol 38 no 8 pp 1188ndash1198 2011
[12] M Ergezer D Simon and D Du ldquoOppositional biogeography-based optimizationrdquo in Proceedings of the IEEE InternationalConference on Systems Man and Cybernetics (SMC rsquo09) pp1009ndash1014 San Antonio Tex USA October 2009
[13] H Ma M Fei D Simon and M Yu ldquoBiogeography-basedoptimization in noisy environmentsrdquo Technique Report 2012httpacademiccsuohioedusimondbbonoisy
[14] M R Lohokare S S Pattnaik B K Panigrahi and S DasldquoAccelerated biogeography-based optimization with neighbor-hood search for optimizationrdquo Applied Soft Computing vol 13no 5 pp 2318ndash2342 2013
[15] W Gong Z Cai and C X Ling ldquoDEBBO a hybrid differentialevolution with biogeography-based optimization for globalnumerical optimizationrdquo Soft Computing vol 15 no 4 pp 645ndash665 2011
[16] X Li J Wang J Zhou and M Yin ldquoA perturb biogeographybased optimization with mutation for global numerical opti-mizationrdquo Applied Mathematics and Computation vol 218 no2 pp 598ndash609 2011
[17] S M Elsayed R A Sarker and D L Essam ldquoMulti-operatorbased evolutionary algorithms for solving constrained opti-mization problemsrdquo Computers and Operations Research vol38 no 12 pp 1877ndash1896 2011
[18] X Li and M Yin ldquoMulti-operator based biogeography basedoptimization withmutation for global numerical optimizationrdquoComputers amp Mathematics with Applications vol 64 no 9 pp2833ndash2844 2012
[19] G Zhu and S Kwong ldquoGbest-guided artificial bee colonyalgorithm for numerical function optimizationrdquo Applied Math-ematics and Computation vol 217 no 7 pp 3166ndash3173 2010
[20] A Hastings and K Higgins ldquoPersistence of transients inspatially structured ecological modelsrdquo Science vol 263 no5150 pp 1133ndash1136 1994
[21] H Ma ldquoAn analysis of the equilibrium of migration models forbiogeography-based optimizationrdquo Information Sciences vol180 no 18 pp 3444ndash3464 2010
[22] X Yao Y Liu and G Lin ldquoEvolutionary programming madefasterrdquo IEEE Transactions on Evolutionary Computation vol 3no 2 pp 82ndash102 1999
[23] P N SuganthanNHansen J J Liang et al ldquoProblemdefinitionand evaluation criteria for the CEC 2005 special sessionon real parameter optimizationrdquo Technical Report NanyangTechnological University and KanGAL Report 2005005 IITKanpur 2005 httpwwwntuedusghomeepnsugan
[24] C H Goulden Methods of Statistical Analysis John Wiley ampSons New York NY USA 2nd edition 1956
[25] S Garcıa A Fernandez J Luengo and F Herrera ldquoAdvancednonparametric tests for multiple comparisons in the design ofexperiments in computational intelligence and data miningexperimental analysis of powerrdquo Information Sciences vol 180no 10 pp 2044ndash2064 2010
[26] Y Wang Z Cai and Q Zhang ldquoEnhancing the search ability ofdifferential evolution through orthogonal crossoverrdquo Informa-tion Sciences vol 185 no 1 pp 153ndash177 2012
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of
Submit your manuscripts athttpwwwhindawicom
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical Problems in Engineering
Hindawi Publishing Corporationhttpwwwhindawicom
Differential EquationsInternational Journal of
Volume 2014
Applied MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Probability and StatisticsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Mathematical PhysicsAdvances in
Complex AnalysisJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
OptimizationJournal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
CombinatoricsHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Operations ResearchAdvances in
Journal of
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Function Spaces
Abstract and Applied AnalysisHindawi Publishing Corporationhttpwwwhindawicom Volume 2014
International Journal of Mathematics and Mathematical Sciences
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Algebra
Discrete Dynamics in Nature and Society
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Decision SciencesAdvances in
Discrete MathematicsJournal of
Hindawi Publishing Corporationhttpwwwhindawicom
Volume 2014 Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014
Stochastic AnalysisInternational Journal of