Summary of Evolutionary Computing

Overview

Last two weeks we looked at evolutionary algorithms.

Overview

This week we are going summaries these into: Basic Principles Applications

Basic Principles 1: Overview

Basic Principles 2: Population

A population of individual possible solutions to a particular problem.

Basic Principles 2: Population

Each individual (or chromosome) encodes the solution.

Basic Principles 2: Population

Each individual needs to evaluated.

Basic Principles 2: Population

Example encoding include: Binary representations Real valued representation

Integers for order based representations.

Basic Principles 3: Reproduction

Parents are selected randomly Better/fitter individual - more likely it is to selected.

Fitness - evaluation individuals

Basic Principles 3: Reproduction

Child produced takes something from both parents.

Basic Principles 3: Reproduction

Different methods of selection are available.

Basic Principles 4: Selection methods: Roulette Wheel Illustration taken from www2.cs.uh.edu/~ceick/ai/EC1.ppt

Fitter the solution-more space on the wheel-more likely to beselected

Best

Worst

Basic Principles 5: Crossover

x amount of ‘genes’ from one parent is included in the child and y amount from the other parent is included.

Basic Principles 5: Crossover

One way to do this is to say: certain point along the chromosome copy Up to this point from one parent

After this point from the other parent.

Crossover causes ‘good’ individuals to combine their ‘genes’ with those of other individuals.

Goal - population of ‘good’ solutions.

combination of different solutions.

speeds up search –average fitness of the population improves rapidly at first.

Basic Principles 6: Mutation Mutation causes random selected changes to an individual.

Basic Principles 6: Mutation Often random valued changes

Basic Principles 6: Mutation

Binary: 11000110 becoming 11010110

Basic Principles 6: Mutation

Real: 2.3 3.4 5.6 becomes 2.3 5.4 5.6

Basic Principles 6: Mutation Low probability event

Basic Principles 6: Mutation Get the population to include different individual solutions.

Basic Principles 7: FitnessEvery individual needs to be evaluated – fitness score.

Basic Principles 7: FitnessThis evaluation is usually in the form of function.

Basic Principles 7: FitnessExamples include:

◦The equation to be solved.

◦Differences between actual and expected results.

Basic Principles 7: FitnessThe only link between the possible solutions and effectiveness to solve the problem.

Basic Principles 8: Population Size.

Need to decide how the population size to managed: Fixed size, maintained by every child added a previous solution is deleted.

Basic Principles 8: Population Size.

Add child without removing individuals?

Replace a small number of individuals each time or the whole population?

Basic Principles 8: Population Size.

Best solution(s) kept in the population – elitism.

Applications 1: Financial/Scheduling

Stock market: http://www.geocities.com/francorbusetti/

mansini.pdf http://www.geocities.com/francorbusetti/

gillikellezi.pdf

Scheduling examples http://www.aridolan.com/ofiles/ga/gaa/Ts

pDemo.aspx

Applications 2: Engineering Assembly

http://www.nait.org/jit/Articles/chen080301.pdf

Biomedical http://www.journals.elsevierhealth.com/p

eriodicals/jjbe/article/PIIS1350453303000213/abstract