Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
11/5/2018
1
INF3490 - Biologically inspired computingSwarm Intelligence, Fuzzy Logic
Weria KhaksarNovember, 06, 2018
11/5/2018
2
206.11.2018
11/5/2018
3
3
Swarm Intelligence: Concept
06.11.2018
Collective behavior emerged from social insectsworking under very few rules.
11/5/2018
4
4
Swarm Intelligence: Concept
06.11.2018
Fish, birds, ants, termites, lions, …
11/5/2018
5
5
Swarm Intelligence: Key Features
06.11.2018
Simple local rules Local interaction Decentralized control Complex global behavior
• Difficult to predict from observing the local rules
• Emergent behavior
11/5/2018
6
6
Swarm Intelligence: General Principles
06.11.2018
Proximity principleThe basic units of a swarm should be capable of simplecomputation related to its surrounding environment.
11/5/2018
7
7
Swarm Intelligence: General Principles
06.11.2018
Proximity principleThe basic units of a swarm should be capable of simplecomputation related to its surrounding environment.
Quality principleApart from basic computation ability, a swarm should be able toresponse to quality factors, such as food and safety.
11/5/2018
8
8
Swarm Intelligence: General Principles
06.11.2018
Proximity principleThe basic units of a swarm should be capable of simplecomputation related to its surrounding environment.
Quality principleApart from basic computation ability, a swarm should be able toresponse to quality factors, such as food and safety.
Principle of diverse responseResources should not be concentrated in narrow region. Thedistribution should be designed so that each agent will be maximallyprotected facing environmental fluctuations.
11/5/2018
9
9
Swarm Intelligence: General Principles
06.11.2018
Proximity principleThe basic units of a swarm should be capable of simplecomputation related to its surrounding environment.
Quality principleApart from basic computation ability, a swarm should be able toresponse to quality factors, such as food and safety.
Principle of diverse responseResources should not be concentrated in narrow region. Thedistribution should be designed so that each agent will be maximallyprotected facing environmental fluctuations.
Principle of stability and adaptabilitySwarms are expected to adapt environmental fluctuations withoutrapidly changing modes since mode changing costs energy.
11/5/2018
10
10
Swarm Intelligence: Particle Swarm Optimization (PSO)
06.11.2018
Mimicking physical quantities such as velocity and positionin bird flocking, artificial particles are constructed to “fly”inside the search space of optimization problems.
11/5/2018
11
11
Swarm Intelligence: Particle Swarm Optimization (PSO)
06.11.2018
Initially, a population of particles is distributed uniformlyin the search space of the objective function of theoptimization problem.
Two quantities are associated with particles, a positionvector 𝑥 and a velocity 𝑣 .
11/5/2018
12
12
Swarm Intelligence: Particle Swarm Optimization (PSO)
06.11.2018
At each time step, the velocities of particles will beupdated according to the following formula:
11/5/2018
13
13
Swarm Intelligence: Particle Swarm Optimization (PSO)
06.11.2018
11/5/2018
14
14
Swarm Intelligence: Particle Swarm Optimization (PSO) Example
06.11.2018PSO algorithm maximizing f(x, y) = -|x**2 - y| (finding squares).
11/5/2018
15
15
Swarm Intelligence: Ant Colony Optimization (ACO)
06.11.2018
The most recognized example of swarm intelligence in realworld is the ants. To search for food, ants will start out fromtheir colony and move randomly in all directions.
11/5/2018
16
16
Swarm Intelligence: Ant Colony Optimization (ACO)
06.11.2018
Once an ant finds food, it returns to colony and leave a trailof chemical substances called pheromone along the path.Other ants can then detect pheromone and follow the samepath.
11/5/2018
17
17
Swarm Intelligence: Ant Colony Optimization (ACO)
06.11.2018
The interesting point is that how often is the path visit byants is determined by the concentration of pheromonealong the path. Since pheromone will naturally evaporateover time, the length of the path is also a factor. Therefore,under all these considerations, a shorter path will befavored because ants following that path keep addingpheromone which makes the concentration strong enoughto against evaporation. As a result, the shortest path fromcolony to foods emerges.
11/5/2018
18
18
Swarm Intelligence: Ant Colony Optimization (ACO)
06.11.2018 Ant Colony Optimization on Traveling Salesman Problem
11/5/2018
19
19
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
Just like ants, bees have similar food collecting behaviors.Instead of pheromones, bees colony optimization algorithmrelies on the foraging behavior of honey bees. At the firststage, some bees are sent out to look for promising foodsources.
11/5/2018
20
20
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
After a good food source is located, bees return back tocolony and perform a waggle dance to spread outinformation about the source. Three pieces of informationare included: (1) distance, (2) direction, (3) quality of foodsource. The better the quality of food source, the morebees will be attracted. Therefore, the best food sourceemerges.
11/5/2018
21
21
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
11/5/2018
22
22
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
11/5/2018
23
23
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
11/5/2018
24
24
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
11/5/2018
25
25
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018
11/5/2018
26
26
Swarm Intelligence: Bee Colony Optimization (BCO)
06.11.2018Using the Bee colony Algorithm to solve the Knight's Tour Problem
11/5/2018
27
27
Swarm Intelligence: Cuckoo Search
06.11.2018
This algorithm is inspired by the brood parasitism behaviorof some species of cuckoo. They will lay their eggs in otherbird's nest. If the host bird find out about this, it will eitherthrow away the intruding egg or simply abandon the wholenest and start a new one. However, some species ofcuckoo are very good at making their eggs the same as thehost's egg, and therefore greatly increase the survivalprobability of their eggs.
11/5/2018
28
28
Swarm Intelligence: Cuckoo Search
06.11.2018
Basic rules of cuckoo search: Each cuckoo lays one egg at a time and dumps it in a
randomly chosen nest. The best nests with high quality of eggs will be brought to the
next generation. The number of host nests is fixed. A host bird will discover the
egg is laid by cuckoo by a probability 𝑃 ∈ 0,1 . The host birdcan get rid of the egg or build a new nest.
11/5/2018
29
29
Swarm Intelligence: Cuckoo Search
06.11.2018
11/5/2018
30
30
Swarm Intelligence: Cuckoo SearchExample: Finding the maximum of 2D Michalewicz' function
06.11.2018
11/5/2018
31
31
Swarm Intelligence: Cuckoo Search
06.11.2018
From left to right: Initial and final positions of the nests marked using dots.
11/5/2018
32
32
Swarm Intelligence: Other Algorithms
06.11.2018
Artificial Immune Systems
Bacterial Foraging
The Shuffled Frog Leaping
The Cat Swarm
Invasive weed optimization
Monkey Search
Water flow-like algorithm
Biogeography-based optimization
The Fish School Search
Bat-inspired Algorithm
Lion Optimization
Firefly algorithm
Dolphin Partner Optimization
Dolphin echolocation algorithm
Flower pollination algorithm
Krill herd
Wolf search
Grey Wolf Optimizer
Water cycle algorithm
The Social spider optimization
Forest Optimization algorithm
...
11/5/2018
33
33
References:
06.11.2018
1. Keerthi, S., Ashwini, K., & Vijaykumar, M. V. (2015). Survey Paper on Swarm Intelligence. International Journal of Computer Applications, 115(5).
2. Keerthi, S., Ashwini, K., & Vijaykumar, M. V. (2015). Survey Paper on Swarm Intelligence. International Journal of Computer Applications, 115(5).
3. Cui, X. Swarm Intelligence.
4. Hu, Y. Swarm intelligence.
5. Bonabeau, E., & Meyer, C. (2001). Swarm intelligence: A whole new way to think about business. Harvard business review, 79(5), 106-115.
11/5/2018
34
3406.11.2018
11/5/2018
35
35
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Fuzzy logic is determined as a set of mathematicalprinciples for knowledge representation based on degreesof membership rather than on crisp membership of classicalbinary logic.
Range of logical values in Boolean and fuzzy logic: (a) Boolean logic; (b) multivalued logic.
11/5/2018
36
36
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
11/5/2018
37
37
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
...
11/5/2018
38
38
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Traditional Control Systems: Need to know detailed physical properties of the system. Most systems are too complex and have to be idealized to
develop a traditional controller. The condition when traditional controllers will work is usually
fairly limited.
Fuzzy logic Control Systems: Do not need much detailed knowledge of the system. If optimization tools are used like GA, can get away with not
knowing much of anything. The condition when fuzzy logic controller will work are much
more robust because they can account for more variability inthe inputs.
11/5/2018
39
39
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Unlike two-valued Boolean logic, fuzzy logic is multi-valued. It deals with degrees of membership and degreesof truth.
Fuzzy logic uses the continuum of logical values between0 (completely false) and 1 (completely true).
Instead of just black and white, it employs the spectrumof colors, accepting that things can be partly true andpartly false at the same time.
11/5/2018
40
40
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Let 𝑋 be a classical (crisp) set and 𝑥 an element. Thenthe element 𝑥 either belongs to 𝑋 𝑥 ∈ 𝑋 or does notbelong to 𝑋 𝑥 ∉ 𝑋 .
Classical set theory imposes a sharp boundary on thisset and gives each member of the set the value of 1, andall members that are not within the set a value of 0.
11/5/2018
41
41
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Who is tall?
11/5/2018
42
42
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Fuzzy membership function
11/5/2018
43
43
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Let 𝑋 be the universe of discourse and its elements be denoted as 𝑥. Inclassical set theory, crisp set 𝐴 of 𝑋 is defined as function 𝑓 𝑥 calledthe Characteristic Function of 𝐴.
11/5/2018
44
44
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Let 𝑋 be the universe of discourse and its elements be denoted as 𝑥. Inclassical set theory, crisp set 𝐴 of 𝑋 is defined as function 𝑓 𝑥 calledthe Characteristic Function of 𝐴. Membership Function
11/5/2018
45
45
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Representation of crisp and fuzzy sets.
11/5/2018
46
46
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Linguistic variables and hedges
11/5/2018
47
47
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
Fuzzy Rules
11/5/2018
48
48
Fuzzy Logic: What is fuzzy thinking?
06.11.2018
How to reason with fuzzy rules?
Evaluating the rule antecedent Implication the results to the consequent
Monotonic Selection
11/5/2018
49
49
Fuzzy Logic: Fuzzy Inference
06.11.2018
Fuzzy Inference Style
Mamdani Inference Style
Sugeno Inference Style
11/5/2018
50
50
Fuzzy Logic: Fuzzy Inference
06.11.2018
Let’s take a look at an example:
Project Funding (adequate, marginal, inadequate) Project Staffing (small, large)
Project Risk (low, normal, high)
11/5/2018
51
51
Fuzzy Logic: Fuzzy Inference
06.11.2018
Let’s take a look at an example:
11/5/2018
52
52
Fuzzy Logic: Fuzzy Inference
06.11.2018
Let’s take a look at an example:
The basic structure of Mamdani-style fuzzy inference
11/5/2018
53
53
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 1: FuzzificationThe first step is to take the crisp inputs, x1 and y1 (project funding andproject staffing), and determine the degree to which these inputs belongto each of the appropriate fuzzy sets.
11/5/2018
54
54
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 2: Rule EvaluationThe second step is to take the fuzzified inputs, 𝜇 0.5, 𝜇 0.2, and𝜇 0.1, and apply them to the antecedents of the fuzzy rules. If a givenfuzzy rule has multiple antecedents, the fuzzy operator (AND or OR) is used toobtain a single number that represents the result of the antecedent evaluation.
11/5/2018
55
55
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 3: Aggregation of the rule outputsAggregation is the process of unification of the outputs of all rules. In otherwords, we take the membership functions of all rule consequents previouslyclipped or scaled and combine them into a single fuzzy set. Thus, the input ofthe aggregation process is the list of clipped or scaled consequent membershipfunctions, and the output is one fuzzy set for each output variable.
11/5/2018
56
56
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 4: DefuzzificationThe last step in the fuzzy inference process is defuzzification. Fuzziness helpsus to evaluate the rules, but the final output of a fuzzy system has to be a crispnumber. The input for the defuzzification process is the aggregate output fuzzyset and the output is a single number.
11/5/2018
57
57
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 4: Defuzzification
How do we defuzzify the aggregate fuzzy set?
There are several defuzzification methods, but probably the most popular one isthe centroid technique. It finds the point where a vertical line would slice theaggregate set into two equal masses. Mathematically this center of gravity(COG) can be expressed as:
11/5/2018
58
58
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 4: Defuzzification
How do we defuzzify the aggregate fuzzy set?
A centroid defuzzification method finds a point representing the center of gravityof the fuzzy set, 𝐴, on the interval, 𝑎𝑏. In theory, the COG is calculated over acontinuum of points in the aggregate output membership function, but inpractice, a reasonable estimate can be obtained by calculating it over a sampleof points. In this case, the following formula is applied:
11/5/2018
59
59
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 4: Defuzzification
How do we defuzzify the aggregate fuzzy set?
11/5/2018
60
60
Fuzzy Logic: Fuzzy Inference
06.11.2018
Step 4: Defuzzification
How do we defuzzify the aggregate fuzzy set?
11/5/2018
61
61
Fuzzy Logic: Some real applications:
06.11.2018Fuzzy Logic Application - Automatic Speed Controller by Unity C#
11/5/2018
62
62
Fuzzy Logic: Some real applications:
06.11.2018NASA | Fuzzy Logic Models for Real-Time Simulations
11/5/2018
63
63
Fuzzy Logic: Prof. Lotfi Zadeh, father of fuzzy logic
06.11.2018Lotfi zadeh, father of mathematical 'fuzzy logic,' died at 96
11/5/2018
64
64
References:
06.11.2018
1. Lee, C. C. (1990). Fuzzy logic in control systems: fuzzy logic controller. I. IEEE Transactions on systems, man, and cybernetics, 20(2), 404-418.
2. Negnevitsky, M. (2005). Artificial intelligence: a guide to intelligent systems. Pearson Education.
3. Klir, G., & Yuan, B. (1995). Fuzzy sets and fuzzy logic (Vol. 4). New Jersey: Prentice hall.
4. Lochan, K., & Roy, B. K. (2015). Control of two-link 2-DOF robot manipulator using fuzzy logic techniques: A review. In Proceedings of Fourth International Conference on Soft Computing for Problem Solving (pp. 499-511). Springer, New Delhi.
5. Petrosino, A., Loia, V., & Pedrycz, W. (Eds.). (2017). Fuzzy Logic and Soft Computing Applications: 11th International Workshop, WILF 2016, Naples, Italy, December 19–21, 2016, Revised Selected Papers (Vol. 10147). Springer.
11/5/2018
65
06.11.2018 65