Lecture 13

Genetic Algorithm
Main steps of GA
Step-1: Coding.
Step-2: Selection (for reproduction).
Step-3: Cross-over.
Step-4: Mutations.
2
Genetic Algorithms: Stochastic Operators
Selection replicates the most successful solutions
found in a population at a rate proportional to their
relative quality. Also known as Reproduction.
Examples of selections schemes:

Roulette wheel selection (probabilistic selection
based on fitness).
Rank selection (pick the best individual each

time).
Tournament selection (select K individuals, and

keep best for reproduction).
3
Genetic Algorithms: Stochastic Operators Cont…
Recombination decomposes two distinct solutions

and then randomly mixes their parts to form novel
solutions. Also known as Crossover.
Mutation randomly perturbs a candidate solution.
4
Genetic Algorithms: Metaphors Cont…
The computer model introduces simplifications

(relative to the real biological mechanisms),
BUT
Surprisingly complex and interesting structures have

emerged out of evolutionary algorithms.
5
Genetic Algorithms: A Simple Genetic Algorithm
Produce an initial population of individuals
Evaluate the fitness of all individuals
while termination condition not met do

select fitter individuals for reproduction
recombine between individuals
mutate individuals
evaluate the fitness of the modified individuals
generate a new population
End while
6
Genetic Algorithms: The Evolutionary Cycle
parents
selection modification
modified
offspring
initiate & evaluation
population
evaluate evaluated off-spring
deleted
members
discard
GA: Example f(x) = x2
Actual count was generated using coin tosses.
Note the similarity with pselect, we obtained what we

expected.
The best strings get more copies, while the weak ones
just die off.
After selection, crossover takes place in two steps
(i) strings combined randomly using coin tosses

(ii) selection of crossing sites using coin tosses.
13
Mating pool after Mate coded New population x value f(x) x2

selection
1 13 0110|1 2 24 1100|0 01100 12 144

2 24 1100|0 1 13 0110|1 11001 25 625
2 24 11|000 4 19
10|011 11011 27 729
3 08
4 19
10|011 01|000 10000 16 256
14
Before After
2000
1500
1000
500
0
Sum Avg Max
15
Notes:
Initial population chosen by four repetitions of five coin tosses
where head = 1 and tail = 0
Reproduction performed through 1 part in 8 simulation of

roulette wheel selection i.e. 3 coin tosses.
Crossover performed by binary decoding of 2 coin tosses (TT

= 002 = 0 = cross site 1, HH = 112 = 3 = cross site 4)
Crossover probability assumed to be unity pc = 1.0.
Mutation probability pm assumed to be 0.001, expected

mutations = 5 x 4 x 0.0001 = 0.02. No mutation expected
during a single generation. None simulated.
18
Visualizing GA (maximizing)
BEFORE
Objective
AFTER
Function
Multidimensional Search Space 19

A Simple Example
“The Gene is by far the most sophisticated program around.”

- Bill Gates, Business Week, June 27, 1994
20
A Simple Example
The Traveling Salesman Problem:
Find a tour of a given set of cities so that

– each city is visited only once
– the total distance traveled is minimized
21
Representation
Representation is an ordered list of city

numbers known as an order-based GA.
1) London 3) Dunedin 5) Beijing 7) Tokyo
2) Venice 4) Singapore 6) Phoenix 8) Victoria
CityList1 (3 5 7 2 1 6 4 8)
CityList2 (2 5 7 6 8 1 3 4)
22
Crossover
Crossover combines inversion and

recombination:
* *
Parent1 (3 5 7 2 1 6 4 8)
Parent2 (2 5 7 6 8 1 3 4)
Child (5 8 7 2 1 6 3 4)
This operator is called the Order1 crossover.

23
Mutation
Mutation involves reordering of the list:
* *
Before: (5 8 7 2 1 6 3 4)
After: (5 8 6 2 1 7 3 4)
24
Genetic Algorithms: Why do they work?
GA offers a means of searching a broad search space.
Different features of problem are represented in search space by

DNA representation
– parallel nature: often many solutions to a problem
– different “good” characteristics represented by particular gene
settings
– some combinations of these genes are better than others
Evolution creates new gene combinations --> new areas of search

space to try.
Key to success: individuals that are fitter are more likely to be

retained and mated; poorer individuals are more likely discarded.
Global search technique, unlike other search techniques that use

heuristics to prune the search space.
Adv Algo: Spring 2006 30

When to use GA?
• Alternate solutions are too slow or overly complicated
• Need an exploratory tool to examine new approaches
• Problem is similar to one that has already been

successfully solved by using a GA
• Want to hybridize with an existing solution
• Benefits of the GA technology meet key problem

requirements
31
Some GA Application Types
Domain Application Types

Control gas pipeline, pole balancing, missile evasion, pursuit
Design semiconductor layout, aircraft design, keyboard

configuration, communication networks
Scheduling manufacturing, facility scheduling, resource allocation
Robotics trajectory planning
Machine Learning designing neural networks, improving classification

algorithms, classifier systems
Signal Processing filter design
Game Playing poker, checkers, prisoner’s dilemma
Combinatorial set covering, travelling salesman, routing, bin packing,

graph colouring and partitioning
Optimization
32
Issues for GA Practitioners
• Choosing basic implementation issues:
– representation
– population size, mutation rate, ...
– selection, deletion policies
– crossover, mutation operators
• Termination Criteria
• Performance, scalability
• Solution is only as good as the evaluation function

(often hardest part)
33
GA: Limitations & Weaknesses
Extremely appealing idea, but falls-short for
practical combinatorial optimization problems
because:
Unnatural to model most applications in terms of genetic

operators like mutation and crossover on bit strings.
Pseudo biology adds another level of complexity between

you and your problem.
Process of selection alone is too systematic and

predictable, not like creativity as we know it.
34
GA: Limitations & Weaknesses
Binary representations limited in operations and for certain

problems alternative operators and representations required.
Real problem info NOT used, hence slow convergence.
Take very long on non-trivial problems due to many

objective function evaluations as compared to classical
optimization techniques.
Analogy with evolution is appropriate, but it took millions of

years to achieve significant improvement. Can we afford to
wait that long?
35

Lecture 13

Uploaded by

Copyright:

Available Formats

Lecture 13

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture 13

Uploaded by

Copyright:

Available Formats

Genetic Algorithm

Step-2: Selection (for reproduction).

Examples of selections schemes:

Rank selection (pick the best individual each

Tournament selection (select K individuals, and

Recombination decomposes two distinct solutions

Mutation randomly perturbs a candidate solution.

The computer model introduces simplifications

Surprisingly complex and interesting structures have

Produce an initial population of individuals

Evaluate the fitness of all individuals

while termination condition not met do

Note the similarity with pselect, we obtained what we

After selection, crossover takes place in two steps

(i) strings combined randomly using coin tosses

Mating pool after Mate coded New population x value f(x) x2

1 13 0110|1 2 24 1100|0 01100 12 144

Reproduction performed through 1 part in 8 simulation of

Crossover performed by binary decoding of 2 coin tosses (TT

Crossover probability assumed to be unity pc = 1.0.

Mutation probability pm assumed to be 0.001, expected

Multidimensional Search Space 19

“The Gene is by far the most sophisticated program around.”

The Traveling Salesman Problem:

Find a tour of a given set of cities so that

Representation is an ordered list of city

Crossover combines inversion and

This operator is called the Order1 crossover.

Mutation involves reordering of the list:

Different features of problem are represented in search space by

Evolution creates new gene combinations --> new areas of search

Key to success: individuals that are fitter are more likely to be

Global search technique, unlike other search techniques that use

Adv Algo: Spring 2006 30

• Need an exploratory tool to examine new approaches

• Problem is similar to one that has already been

• Want to hybridize with an existing solution

• Benefits of the GA technology meet key problem

Domain Application Types

Design semiconductor layout, aircraft design, keyboard

Robotics trajectory planning

Machine Learning designing neural networks, improving classification

Game Playing poker, checkers, prisoner’s dilemma

Combinatorial set covering, travelling salesman, routing, bin packing,

• Solution is only as good as the evaluation function

Unnatural to model most applications in terms of genetic

Pseudo biology adds another level of complexity between

Process of selection alone is too systematic and

Binary representations limited in operations and for certain

Real problem info NOT used, hence slow convergence.

Take very long on non-trivial problems due to many

Analogy with evolution is appropriate, but it took millions of

You might also like