12

Peter Spijker
Technische Universiteit Eindhoven Department of Biomedical Engineering Division of Biomedical Imaging and Modeling

13

Basics of Genetic Algorithms and some possibilities

California Institute of Technology Materials Process and Simulation Center Biochemistry & Molecular Biophysics

November 25, 2003

Presentation Overview Purpose of presentation General introduction to Genetic Algorithms (GAs) Biological background
Origin of species Natural selection

13

Genetic Algorithm
Search space Basic algorithm Coding

Methods
Examples

Possibilities

Purpose of presentation Optimising parameters of force fields is a difficult and time consuming task Use of optimising methods might be of use Methods:
- steepest descent - simulated annealing (Monte Carlo)

13

- genetic algorithms

Brief introduction to genetic algorithms in lecture style

General Introduction to GAs

13

Genetic algorithms (GAs) are a technique to solve problems which need optimization GAs are a subclass of Evolutionary Computing GAs are based on Darwins theory of evolution

History of GAs Evolutionary computing evolved in the 1960s. GAs were created by John Holland in the mid-70s.

Biological Background (1) The cell Every animal cell is a complex of many small factories working together The center of this all is the cell nucleus The nucleus contains the genetic information

13

Biological Background (2) Chromosomes

13

Genetic information is stored in the chromosomes Each chromosome is build of DNA Chromosomes in humans form pairs There are 23 pairs The chromosome is divided in parts: genes

Genes code for properties

Every gene has an unique position on the chromosome: locus

Biological Background (3) Genetics

13

The entire combination of genes is called genotype A genotype develops to a phenotype Alleles can be either dominant or recessive Dominant alleles will always express from the genotype to the fenotype Recessive alleles can survive in the population for many generations, without being expressed.

Biological Background (4) Reproduction Reproduction of genetical information

13

Mitosis is copying the same genetic information to new offspring: there is no exchange of information

Mitosis is the normal way of growing of multicell structures, like organs.

Biological Background (5) Reproduction Meiosis is the basis of sexual reproduction After meiotic division 2 gametes appear in the process In reproduction two gametes conjugate to a zygote wich will become the new individual Hence genetic information is shared between the parents in order to create new offspring

13

Biological Background (6) Reproduction During reproduction errors occur Due to these errors genetic variation exists Most important errors are: Recombination (cross-over) Mutation

13

Biological Background (7) Natural selection

13

The origin of species: Preservation of favourable variations and rejection of unfavourable variations. There are more individuals born than can survive, so there is a continuous struggle for life. Individuals with an advantage have a greater chance for survive: survival of the fittest.

Biological Background (8) Natural selection Important aspects in natural selection are: adaptation to the environment

13

isolation of populations in different groups which cannot mutually mate If small changes in the genotypes of individuals are expressed easily, especially in small populations, we speak of genetic drift Mathematical expresses as fitness: success in life

Presentation Overview Purpose of presentation General introduction to Genetic Algorithms (GAs) Biological background
Origin of species Natural selection

13

Genetic Algorithm
Search space Basic algorithm Coding

Methods
Examples

Possibilities

Genetic Algorithm (1) Search space Most often one is looking for the best solution in a specific subset of solutions

13

This subset is called the search space (or state space) Every point in the search space is a possible solution Therefore every point has a fitness value, depending on the problem definition GAs are used to search the search space for the best solution, e.g. a minimum
2.5 2

1.5

Difficulties are the local minima and the starting point of the search

0.5

100

200

300

400

500

600

700

800

900

1000

Genetic Algorithm (2) Basic algorithm

13

Starting with a subset of n randomly chosen solutions from the search space (i.e. chromosomes). This is the population This population is used to produce a next generation of individuals by reproduction Individuals with a higher fitness have more chance to reproduce (i.e. natural selection)

Genetic Algorithm (3) Basic algorithm Outline of the basic algorithm

13

: Create random population of n chromosomes

FITNESS : Evaluate fitness f(x) of each chromosome in the population NEW POPULATION 0 SELECTION : Based on f(x)

1 RECOMBINATION : Cross-over chromosomes 2 MUTATION 3 ACCEPTATION 3 : Mutate chromosomes : Reject or accept new one

REPLACE : Replace old with new population: the new

generation
4 5 TEST LOOP : Test problem criterium : Continue step 1 4 until criterium is satisfied

Genetic Algorithm (4) Coding

13

Normal cells are diploid (containing 2 complete sets of chromosomes) On the contrary gametes are haploid Formalizing diploid reproduction is much more difficult than haploid Diploid populations have an extra dimension compared to haploid populations For simplicity therefore only haploid genetic algorithms

Genetic Algorithm (5) Coding Chromosomes are encoded by bitstrings

13

Every bitstring therefore is a solution but not necisseraly the best solution The way bitstrings can code differs from problem to problem

1 0 0 1

Either: sequence of on/off or the number 9

Genetic Algorithm (6) Coding Recombination (cross-over) can when using bitstrings schematically be represented:

13

1 0 0 1 1 0 1

0 1 0 1 1 1 0

1 0 0 1 1 1 0

0 1 0 1 1 0 1

Using a specific cross-over point

Genetic Algorithm (7) Coding

13

Mutation prevents the algorithm to be trapped in a local minimum In the bitstring approach mutation is simpy the flipping of one of the bits

1 0 0 1 1 0 1

1 1 0 1 1 0 1

Genetic Algorithm (8) Coding

13

Both recombination and mutation depend a lot on the exact definition of the problem and the choice of representing the chromosomes (e.g. no bitstrings) Different encodings can be used:
Binary encoding

Permutation encoding
Value encoding Tree encoding

Focus in this presentation stays with binary encoding

Example Minimum of Function (1) First example shows how to find the minimum of a function
2.5

13

1.5

0.5

Minimum f(x) at x = 809

1100101001

Example Minimum of Function (2)

13

1.8

1.6

1.4

1.2

Mean fitness
1.3

Best fitness

0.8

0.6

0.4 Generation 1 0.2 0 200 400 600 800 1000

Best Fitness Mean Fitness 1.2

1.1 1200 1

Individual

Best individual
Fitness

0.9

0.8

0.7

0.6

0.5

0.4

0.3

10

20

30

Generations

40 Generations

50

60

70

80

Example Minimum of Function (3) Interactive show of this algorithm with Matlab Using the function: genalg2() Variables:
Population size
Bitstringlength Mutation chance Recombination chance Starting population adaption

13

Genetic Algorithm (9) Remarks

13

It is clear from the example that the convergence speed of the algorithm depends on many factors:
Population size

Mutation probability
Recombination probability Elitism

Selection methods
Random selection of parents Roulette wheel selection of parents

Strong point GAs: mutation prevents from falling in a local minimum, recombination initiates a fast first convergence

Example Checkboard (1) We are given an n by n checkboard in which every field can have a different colour from a set of four colours.

13

Goal is to achieve a checkboard in a way that there are no neighbours with the same colour (not diagonal)

10 1 2 3 4 5 6 7 8 9 10

10 1 2 3 4 5 6 7 8 9 10

Example Checkboard (2) Chromosomes represent the way the checkboard is coloured. Chromosomes are not represented by bitstrings but by bitmatrices The bits in the bitmatrix can have one of the four values 0, 1, 2 or 3, depending on the colour

13

Crossing-over involves matrix manipulation instead of point wise operating. Crossing-over can be combining the parential matrices in a horizontal, vertical, triangular or square way
Mutation remains bitwise changing bits in either one of the other numbers

Example Checkboard (3) Fitnesscurve for the checkboard example

13

170

165

160
Fitness

155

150

145

140

Best Fitness Mean Fitness

135

130

100

200

300 Generations

400

500

600

This problem can be seen as a graph with n nodes and (n-1) edges, so the fitness f(x) is easily defined as: f(x) = 2 (n-1) n

Example Checkboard (4) Fitnesscurves for different cross-over rules

13

Fitness

160 150 140 130

100

200

300

400

500

200

400

600

800

Horizontal Cutting Crossing Over 180 170 180 170 160 150 140 130

Verical Cutting Crossing Over

Fitness

160 150 140 130

200

400 Generations

600

800

500 1000 Generations

1500

Example Checkboard (5) Interactive show of this algorithm with Matlab Using the functions:
main() checkers() bestindividual()

13

mutate()
recombine() select() showbestindividual()

Possibilities Using the genetic algorithm to optimise parameters for a force field Parameters are real numbers, so adaptations of these algorithms is required Value incoding vs. bitstring encoding Difficulties:
Definition fitness function Integration algorithm with software

13

Further Questions

13