paper review for engg6140 memetic algorithms

35
1 Paper Review for ENGG6140 Memetic Algorithms By: Jin Zeng Shaun Wang School of Engineering University of Guelph Mar. 18, 2002

Upload: bambi

Post on 06-Jan-2016

64 views

Category:

Documents


1 download

DESCRIPTION

Paper Review for ENGG6140 Memetic Algorithms. By: Jin Zeng Shaun Wang School of Engineering University of Guelph Mar. 18, 2002. Contents. Introduction MA and GA Basic MA Examples Conclusions. Introduction History of MA. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Paper Review for ENGG6140 Memetic Algorithms

1

Paper Review for ENGG6140

Memetic Algorithms

By: Jin Zeng

Shaun Wang

School of Engineering

University of Guelph

Mar. 18, 2002

Page 2: Paper Review for ENGG6140 Memetic Algorithms

2

ContentsContents

Introduction MA and GABasic MAExamplesConclusions

Page 3: Paper Review for ENGG6140 Memetic Algorithms

3

IntroductionIntroductionHistory of MAHistory of MA

‘Meme’: word introduced by Richard Dawkins when he describe cultural evolution in his best-seller book “The Selfish Gene’’ (‘76).

“Memetic Algorithms’’ Analogous role of gene but in the field of cultural evolution.‘Memetic Algorithms’ , firstly proposed by P. Moscarto. (‘89)

MA has been widely applied in optimization and solving many NP hard problems successfully.

Page 4: Paper Review for ENGG6140 Memetic Algorithms

4

Introduction Introduction What is ‘Meme’?What is ‘Meme’?

Meme is the basic unit of cultural transmission, in analagy to gene in genetic transmission.

Meme is replicated by imitation. It can be changed by the owner for adaption. Examples: ideas, clothing fashion and NBA. High-extent variation occurs in cultural

transmission.

Page 5: Paper Review for ENGG6140 Memetic Algorithms

5

IntroductionIntroductionCultural EvolutionCultural Evolution

When a meme passed between individuals, the individual will adapt the meme as it sees best.

Shared characteristics are not inherited due to simple processes of recombination of previous solutions

Using historical information and an external logic to speed-up the process.

Page 6: Paper Review for ENGG6140 Memetic Algorithms

6

Introduction Introduction What is MA?What is MA?

MA mimics the process of cultural evolution Characterization of evolutionary algorithms that

can hardly fit the GAs methaphor - no, or small, relation with biology

‘Hybrid GAs’ MAs ‘Scatter Search’ (Glover, ‘77) MAs

Page 7: Paper Review for ENGG6140 Memetic Algorithms

7

IntroductionIntroductionWhy MA?Why MA?

In general, there are two ways to searching the solution space: Exploration: Investigate the new and unknown

areas in the search space; Exploitation: Make use of knowledge found

before to help find better solutions Both are necessary but contradictory in solving an

optimization problem.

Page 8: Paper Review for ENGG6140 Memetic Algorithms

8

IntroductionIntroduction Why MA? (cont.)Why MA? (cont.)

The limitation of former algorithms: GA: using parallel searching technique.

Good at avoiding local optima Not well suited for finely tuned search.

LS: improvement heuristics. Find local optima quickly. Highly depending on the start point. Hard to find a global optimum.

Page 9: Paper Review for ENGG6140 Memetic Algorithms

9

IntroductionIntroduction Why MA? (cont.)Why MA? (cont.)

Combination of GA + Local SearchMA

GA: For exploration; LS: For exploitation; Result: higher efficiency and better effect.

Page 10: Paper Review for ENGG6140 Memetic Algorithms

10

IntroductionIntroductionCombination MethodsCombination Methods

Two kinds of Combinations:

Baldwin Effect Based Lamarkian Evolution Based

LS is used to modify the structure of the problem. The improvement is not inherited by the children.

The improvement of LS will be inherited in the children.

Wrong in biological evolution. But effective in optimization.

Page 11: Paper Review for ENGG6140 Memetic Algorithms

11

MA and GAMA and GASimilaritiesSimilarities

Both MA and GA model an evolutionary process. Both MA and GA have the process of

generalization, recombination (crossover) and mutation. Some changes occur in the process.

Both MA and GA use fitness function to evaluate the changes in the process thus both of them are applied in optimization successfully.

Page 12: Paper Review for ENGG6140 Memetic Algorithms

12

MA and GAMA and GADifference Difference

MA GA Models Cultural Evolution Bio Evolution Basic Unit Meme Gene Flow Process Information Bio Characteristics Evolution Speed Fast Slow Copying Fidelity Low High Mutation Rate High Low

Page 13: Paper Review for ENGG6140 Memetic Algorithms

13

Basic MABasic MAFlow Chart ProcessFlow Chart Process

Page 14: Paper Review for ENGG6140 Memetic Algorithms

14

Basic MABasic MAPseudo Code of MAPseudo Code of MA

procedure Memetic Algorithm(); begin Generalization(); repeat Crossover(); Mutation(); P := select (P); if P converged then );(: PdLSMutationAnP until terminate=true; end;

Page 15: Paper Review for ENGG6140 Memetic Algorithms

15

Basic MABasic MAGeneralizationGeneralization

Generalization ( ) begin for 1:j to popsize do ();: lutionGenerateSoi );(: ihLocalSearci Add individual i to P endfor end

Page 16: Paper Review for ENGG6140 Memetic Algorithms

16

Basic MABasic MACrossoverCrossover

Crossover ( ) begin for 1:i to crossover# do Select two parents Pii ba , randomly; );,(: bac iiCrossoveri );(: cc ihLocalSearci

Add individual ci to P ; endfor end

Page 17: Paper Review for ENGG6140 Memetic Algorithms

17

Basic MABasic MAMutationMutation

Mutation ( ) begin for 1:i to mutations# do Select and individual Pi randomly );(: iMutationim );(: mm ihLocalSearci

Add individual mi to P ; endfor end

Page 18: Paper Review for ENGG6140 Memetic Algorithms

18

Basic MABasic MALocal SearchLocal Search

Full Local Search and Partial Local Search Demo of FLS

Y Original Solution

Solution after Recombination or Mutation

Solution after Local Search

X

Page 19: Paper Review for ENGG6140 Memetic Algorithms

19

Basic MABasic MADemonstration of MADemonstration of MA

Example Problems: Y= f(x);• Parameters of MA:• Population: 5;• Xover rate:0.4; (# of Xover: 5x0.4=2)• Mutation rate: 0.4; (# of Mutation: 5x0.4=2)• Local Search: Full

Page 20: Paper Review for ENGG6140 Memetic Algorithms

20

Basic MABasic MADemonstration of MA (Continued)Demonstration of MA (Continued)

Y

X

Random Generalized Solution

Solutions After Local Search

X

Solutions After Mutation

Solutions After Crossover

Solutions After Local SearchY

A. Generalization and LS

B. Crossover and Mutation

Page 21: Paper Review for ENGG6140 Memetic Algorithms

21

Basic MABasic MADemonstration of MA (Continued)Demonstration of MA (Continued)

Y

X

Solutions After Crossover

X

Y

Solutions After Local SearchSolutions After CrossoverSolutions After MutationSolutions After Local Search

C. Local Search after Crossover and Mutation

D. Population Selection

Page 22: Paper Review for ENGG6140 Memetic Algorithms

22

Basic MABasic MAEffect of Crossover and MutationEffect of Crossover and Mutation

Both can be used for exploring the search space by “jumping” to new regions to start new local search;

Crossover

Searching the region between two or more specified points; Mutation

Searching the undirected region randomly;

Page 23: Paper Review for ENGG6140 Memetic Algorithms

23

Basic MABasic MAAdvantage of MAAdvantage of MA

Combining the advantages of GA and LS while avoid the disadvantages of both;

GA ensures wide exploration in the solution space Through local search, the space of possible

solutions can be reduced to the subspace of local optima.

When the scale of problem increases, the advantages becomes remarkable.

Page 24: Paper Review for ENGG6140 Memetic Algorithms

24

Basic MABasic MADisadvantage of MADisadvantage of MA

The proportion of computations used in exploration and exploitation depends on the real optimization problem.

It is hard to determine the best depth of local search,.

Page 25: Paper Review for ENGG6140 Memetic Algorithms

25

MA ExamplesMA ExamplesSome Implementation Examples of MASome Implementation Examples of MA

Quadratic Assignment Problem (QAP) Traveling Salesman Problem (TSP) Vehicle Routing Graph Partitioning Scheduling The Knapsack Problem

Page 26: Paper Review for ENGG6140 Memetic Algorithms

26

MA ExamplesMA ExamplesApply Local Search to MA in QAPApply Local Search to MA in QAP

For any permutation solution being explored, the procedure for the local search be executed once or several times –– partial local searchpartial local search (PLS) (PLS)

The procedure for the local search be repeated many times until no further improvement is possible –– full local searchfull local search (FLS) (FLS)

Page 27: Paper Review for ENGG6140 Memetic Algorithms

27

MA ExamplesMA ExamplesDerived Two different MAs for QAPDerived Two different MAs for QAP

PGAPGA –– starts with an initial population of randomly generated individuals. For each individual, after xover and mutation, a PLS is performed.

FGAFGA –– relies on FLS, full local search are carried out on all individuals at the beginning and at the end of a SGA run.

Page 28: Paper Review for ENGG6140 Memetic Algorithms

28

MA ExamplesMA ExamplesBriefly Steps involved for the PGABriefly Steps involved for the PGA

The steps for PGA is same as the Basic MA. The procedures for the local search only executed

once or several times after each xover and mutation.

Page 29: Paper Review for ENGG6140 Memetic Algorithms

29

MA ExamplesMA ExamplesBriefly Steps involved for the FGABriefly Steps involved for the FGA

1. Randomly generate an initial population. Perform FLS on each individual.

2: While terminating criterion is not reached, continue with procedures as spelled out for the SGA.

3: Perform FLS on the best solution and output the final solution.

Page 30: Paper Review for ENGG6140 Memetic Algorithms

30

MA ExamplesMA ExamplesComparison of FGA and PGAComparison of FGA and PGA

The effectiveness of FLS depends on the starting solution and the exchange routine.

PLS can be carried out more frequently, the algorithm is therefore able to spread out the search by exploring many small-localized regions, thus reducing the likelihood of the algorithms being trapped in a local optimum.

Page 31: Paper Review for ENGG6140 Memetic Algorithms

31

MA ExamplesMA ExamplesComparison of FGA and PGA (cont.)Comparison of FGA and PGA (cont.)

As the size of the problem scales up, it is difficult to carry out FLS freely due to its great computational intensity.

PLS is carried out for almost all the individuals in addition to the SGA evolutionary mechanisms, the capability of the SGA in evolving towards fitter individuals is greatly enhanced.

Page 32: Paper Review for ENGG6140 Memetic Algorithms

32

MA ExamplesMA ExamplesComparison of FGA and PGA (cont.)Comparison of FGA and PGA (cont.)

FLS limits the exploratory capability of the SGA, it will reduce the chance of the FGA reaching the global optimum.

PGA has a greater chance of obtaining the global optimum as compared to FGA.

Page 33: Paper Review for ENGG6140 Memetic Algorithms

33

MA ExamplesMA ExamplesComparison of a typical run on problem Els19 Comparison of a typical run on problem Els19

for SGA, PGA and FGAfor SGA, PGA and FGA

Average Cost

Generation1.6x107

1.8x107

2.0x107

2.2x107

2.4x107

2.6x107

2.8x107

3.0x107

10 110 210 310 410

-Optimum

-FGA-PGA

-SGA

Page 34: Paper Review for ENGG6140 Memetic Algorithms

34

ConclusionConclusion

MA provides a more efficient and more robust way to the optimization problem.

MA combines global and local search by using EA to perform exploration while another local search method performs exploitation.

MA can solve some typical optimization problem where other meta-heuristics have failed.

Page 35: Paper Review for ENGG6140 Memetic Algorithms

35

Thank you!Thank you!