1568-4481-1-SP

Embed Size (px)

Citation preview

  • 8/12/2019 1568-4481-1-SP

    1/4

    AbstractRecently, Engineering has complicated limits in manyfields. Large-scale optimization problems are very hard to get an

    optimum solution in short arithmetic time. In this case, we use

    evolutionary algorithm as a main mechanism that can solve theproblems in short arithmetic time. The evolutionary Algorithm is used

    as an important mechanism that can solve these problems in suitable

    arithmetic time. Mask's creation has become a basis for DNA

    computing using GA(Genetic Algorithm); it makes a high propensity

    for randomization and suitable masks. However, software simulation

    requires much simulation time. In this paper, we propose a hardware

    mask creation system for DNA computing using GA and compare it

    with a software model.

    KeywordsDNA Computing, Genetic Algorithm, Mask, Randomnumber generator

    I. INTRODUCTION

    NA computing has problems. It takes a long time to make

    good compatibility masks with GA. For better

    compatibility, we need to use a genetic algorithm that is an

    evolutionary algorithm to apply DNAC. By doing so, we can

    speed up the rates of generation. Further, since hardware

    implementation of the genetic algorithm will solve the problem

    of the generation of mask that needs to be made an executed in

    real-time, it is important that the genetic algorithm needs to

    execute repeatedly to extend over several generations.

    There are two problems when we simulate software

    implementation for DNAC. The process speed is slow because

    it needs to compare all of the data. We can solve this problem by

    using hardware implementation that uses parallel processing.This speeds up the process. We use the process to execute

    simple counts perfectly. But it is hard to use parallel processing

    Sang-Seol Lee, Jae-Yeon Song, Kyu-Yeul Wang, Byung-Soo Kim,

    Jae-Young Choi, Seong-Seob Shin are with the Inha University, Incheon, South

    Korea ( e-mail : [email protected], o [email protected],

    [email protected], [email protected], [email protected],

    [email protected], ).

    Dong-Sun Kim is with the Department of Advanced Mobile Technology

    Research Center, Korea Electronics Technology Institute, Korea. (e-mail:

    [email protected]).

    Duck-Jin Chung is currently a professor at the school of Information and

    Communication Engineering of INHA University, Incheon, Korea.

    (corresponding author to provide phone: +82-32-874-1663; e-mail:[email protected]).

    when we need to implement huge DNA with GA. When we

    make the mask using the random method, it causes some parts of

    the mask to have a defect. We aim to solve this problem. We use

    the genetic algorithm in the way of the survival of the fittest.

    Therefore, we can solve the problems and improve this

    processing.

    In this paper, we simulate genetic algorithm by using Visual

    studio to get the result about performance improvement. We

    implement genetic algorithm by using Verilog-HDL and C

    programming language to know speed rates. The rest of the

    paper is organized as follows. We describe the genetic

    algorithm in terms of the definition and the requirements for

    parallel processing. Also, in section 2, the proposed genetic

    algorithm architecture is discussed in terms of a random number

    generator module, the way of selection, crossover block,

    mutation block, and the hardware implementation for

    compatible function blocks. In section 3, we present the resultthat is compatible with the software simulation and the hardware

    simulation in the genetic algorithm. The paper concludes with a

    discussion of DNA computing using the genetic algorithm in

    section 4.

    II. IMPLEMENTATION OF GAFOR MASK CREATION

    A. What is the Genetic algorithm?

    The genetic algorithm has individual of high fitness. The high

    quality of individual will survive more than others. This optimal

    algorithm is based on law of the natural selection. It is widelyused in area of pattern recognition, neural network, voice

    recognition and robotics. The genetic algorithm was introduced

    by John Holland. The algorithm encodes a lot of individual

    characteristics such as genetic organization. After that, it

    evolves a parallel way using a genetic operator. Therefore, the

    parallel way uses the techniques of probability search to reach

    the optimized state.

    Figure.1 depicts a flow chart of the genetic algorithm. It

    describes the procedure. The many masks represent the type of

    chromosome and then the population of generated

    chromosomes initialize. The randomly initialized individuals

    are evaluated by the algorithms problem-solving capability.

    The individuals' fitness is decided through this evaluation

    Mask Creation for DNA Computing

    Using Genetic Algorithm

    Sang-Seol Lee, Jae-Yeon Song, Kyu-Yeul Wang, Byung-Soo Kim, Jae-Young Choi, Seong-Seob Shin,

    Dong-Sun Kim and Duck-Jin Chung

    D

    World Academy of Science, Engineering and Technology 50 2009

    289

    http://www.pdfonline.com/easypdf/?gad=CLjUiqcCEgjbNejkqKEugRjG27j-AyCw_-AP
  • 8/12/2019 1568-4481-1-SP

    2/4

    process. When genes generate children, the survival of the fittest

    is used. Individuals are selected through a fitness-based process,

    where fitter solutions are typically more likely to be selected.

    This selection procedure is repeated until it reaches the end

    condition

    Fig. 1 the flow chart about genetic algorithm

    B. Parallel processing

    The genetic algorithm is the principle of the naturalprogression of evolution. GA makes population to evolve

    continuously and the population converges into optimum

    solution. The children are produced by genetic operation:

    selection, crossover, mutation and reproduction. It is a simple

    theory and easy to apply. GA has a drawback, which is that it

    makes the process too slow when it is simulated by software in

    computers. We need parallel processing that runs evolutionary

    arithmetic using different several groups to overcome the speed

    problem.

    C. Operation

    The random number generator generates random variables to

    make the mask in each address. We calculate the fitness of themask to save a memory. Among the masks, we choose the best

    mask to have high fitness. After that, the parent generation is

    initialized. We generate each address by a random number

    generator. Each address aims at relevant data A and data B. The

    data sets are parent A and parent B. The crossover chooses a

    random point that is generated by a random number generator to

    exchange genetic traits of parent A and B. We operate the

    mutation module. It depends on the probability of mutation. We

    got the result that the mask is convergence from one hundred

    individuals..

    TABLE I

    THE PSEUDO-CODE OF THE PROPOSED GA FOR MASK CREATION

    functionGA for Mask Creation(population, FITNESS) returnsan

    individual

    input:population, a set of individuals

    FITNESS, a function that measures the fitness of an

    individual

    repeat

    Operation parameter random number generator

    parentsSELECTION(population,FITNESS)

    population Crossover & Mutation (parents)

    untilsome individual is fit enough

    returnthe best individual inpopulation, according to FITNESS

    Fig. 2 the flow chart about genetic algorithm

    D. Selection

    The selection module selects the object of parents to make

    two child buffers. Data transmission consists of handshaking

    protocol between memory and the random number generator.

    The right block selects the object of parents for crossover and

    mutation. The left block has the whole control. The roulette

    wheel selection process is used in our mask creator. As shown in

    Fig. 3, after it selects two parents in memory, and then the

    parents pass to two child buffer in one clock, it contributes to

    both utilized parents for fast arithmetic and reduction of

    hardware size.

    .

    Buffer1-1 Buffer1-2

    Random value

    Generator

    crossover point

    Mutation rate

    Mutation point

    Mutation change

    value

    Buffer2-1 Buffer2-2

    Mutation

    Controller

    Crossover Controller

    Parent Buffer

    Fig. 3 operation block

    World Academy of Science, Engineering and Technology 50 2009

    290

    http://www.pdfonline.com/easypdf/?gad=CLjUiqcCEgjbNejkqKEugRjG27j-AyCw_-AP
  • 8/12/2019 1568-4481-1-SP

    3/4

    E. Crossover & Mutation module

    This module is designed so that two operations may be

    performed continuously. In case of crossover, it uses 1- point

    crossover method. Mutation uses a single point mutation

    method. Crossover and mutation are combined. As shown in Fig.

    3, the block is hardware structure of crossover and mutation.

    Parents receive the intersection points in the random number

    generator; each pair undergoing crossover and mutation. The

    mechanics of crossover and mutation are simple, involving the

    random number generator. Nonetheless, this module is efficient

    in optimization.

    F. Fitness function module

    To test the fitness function for DNA computing in masks, we

    use decision block with digit recognition.[11]

    We compare the original train data and masks. If the train

    data and mask are the same, we increase the fitness value and

    store it together with mask in memory.

    Fig 4. Decision Block & Fitness

    G. Selection bits (Mask bits)

    In this architecture, selection bits and the number of parallel

    processing units are 31 and 1000. In software and hardware

    simulations, we archived mask best accuracy on 31 bits (fig 5).

    Fig 5. Accuracy according bits size

    TABLE II

    FUNCTION PROPERTIES

    Function Quantity

    Selection Roulette wheel

    Crossover One-point crossover

    Mutation Single port mutation , 0.05# of mask bits 31 bits

    Modified Gollmann casecade(HW)Random

    number

    generatorrPseudo random number(SW)

    III. R ESULTS

    The aim of this paper is a mask creation for DNA computing

    using genetic algorithm. The fitness of the first mask is shown in

    Fig.6 We calculate the value of fitness from the random number

    generator and then we save the value of fitness to the memory.

    We repeat this process 100times. It is the first generated mask.

    The function's fitness is large at GA. Therefore, we need more

    time to create the mask by using GA than by using the ordinary

    way.

    We simulated this system by using software. At that time, we

    needed over 21 hours to simulate the system by using a

    computer that is dual core2 6600, and ram 2G. On the other

    hand, we can get a better result by using hardware simulation. In

    this case, we just need about 200 seconds to simulate The

    system on the same computer system. It improves the processing

    time by 380 times. Also, we ensure the result of more than

    99.8% accuracy from self-matching and about 93% accuracy

    from test pattern matching by using DNA computing.

    Fig 6. 1st mask fitness

    Fig 7. The matching result simulated by ModelSim.

    World Academy of Science, Engineering and Technology 50 2009

    291

    http://www.pdfonline.com/easypdf/?gad=CLjUiqcCEgjbNejkqKEugRjG27j-AyCw_-AP
  • 8/12/2019 1568-4481-1-SP

    4/4

    Fig. 8. Accuracy results

    IV. CONCLUSION

    In this paper, we can save mask generation time by using GA

    that implements parallel processing and hardware

    implementation. We can apply this system to real-time because

    the proposed way has decreased the calculation time and the

    number of iterations of genetic algorithm. In addition, we can

    put trains in this system when we need to use a new pattern. We

    can get good results from the proposed system.

    REFERENCES

    [1] J. H. Holland, "Adaptation in Natural and Artificial Systems," Univ. of

    Michigan Press, Ann Arbor, 1975.

    [2] Jenkins, Robert J. "ISAAC". http://burtleburtle.net/bob/rand/isaac.html

    [3] Melanie Mitchell, An Introduction to Genetic Algorithm, 1997

    [4] Myung-Sook Ko, Joon-Min Gil Acellular Learning Strategy for Local

    Search in Hybrid Genetic Algorithms, 2001

    [5] D. E. Goldberg, ed, Genetic Algorithm in search, Optimization,andMachine Learning, Addison-Wesley, 1989.

    [6] K. Dejong, An analysis of behavior of a class of genetic adaptive

    system, Ph.D Thesis, University of Michigan,1975.

    [7] M. Srinivas, and L. M. Patnaik, Adaptive Probabilities of Crossover and

    Mutation in Genetic Algorithm, IEEE Transactions on System, Man

    And Cybernetics, vol. 24, no.4, pp. 656667, April 1994.

    [8] L. B. Booker, D. E. Goldberg, and J. H. Holland, Classifier Systems and

    Genetic Algorithms, Technical Report, No. 8, University of Michigan,

    1987.

    [9] Philips Thrift, Fuzzy Logic Synthesis with Genetic

    Algorithms,Proceedings of the Fourth international Conference on

    Genetic Algorithms, 1991.

    [10] Byung-Tak Zhang and Joo-Kyung Kim, DNA Hypernetworks for

    Information Storage and retrieval, Lecture Notes in Computer Science

    (DNA12), LNCS 4287, pp.298-307, 2006.

    [11] Byung-Soo Kim, Joo-Kyung Kim, Oh Hyuk Kwon, Seung Kon Hwang,Jae-Yeon Song, Byoung-Tak Zhang, Chong Ho Lee, Jaehyun Park, and

    Duck-Jin Chung, Hardware Implementation of the pattern recognize

    processor base on DNA cimputiong technique, ITC-CSCC 2007, 2007

    World Academy of Science, Engineering and Technology 50 2009

    292

    http://www.pdfonline.com/easypdf/?gad=CLjUiqcCEgjbNejkqKEugRjG27j-AyCw_-AP