Genetic Algorithms

Various Operators

Sabah Sayed
Department of Computer Science
Faculty of Computers and Artificial Intelligence
Cairo University
Egypt

3/8/2024 1
 Remember: Mechanism Of GAs
Begin

Initialize population

T =0

Evaluate Solutions

Optimum N
Solution?
Selection
Y

T=T+1 Stop Crossover

Mutation

Replacement

2
 Various Strategies for the Genetic Operators

• Different GAs use different ……… strategies.

– Representation (encoding/decoding)
– Crossover
– Mutation
– Selection
– Replacement
 Various Representations
Character Array?
String Array?
Floating Point?
Integer?
Permutation?

Phenotype space Genotype space =

Encoding {0,1}L
(representation) 10010001
10010010
010001001
011101001
Decoding
(inverse representation)
3/8/2024 4
English Word Generation Example
Objective fn
Problem: Generate valid English word of
length 5 characters
Crossover

t t w w Dictionary
h h o o
r f r f
e s e s
s . . . e . . s . . . . e
Chr 1 Chr N
English Word Generation Example
Problem: Generate valid English word of Objective fn

length 5 characters
Mutation

t t w w Dictionary
h h o o
r e r f
e s d s
s . . . e . . s . . . . e
Chr 1 Chr N
 Integer Representation
• Some problems naturally have integer variables,
e.g. image processing parameters

• Others take categorical values from a fixed set e.g.

{blue, green, yellow, pink}

• N-point / uniform crossover operators work

• Extend bit-flipping mutation to make

• Random choice (categorical variables)

3/8/2024 7
 Uniform Crossover for Integers
• Assign 'heads' to one parent, 'tails' to the other
• Flip a coin for each gene of the first child
• Make an inverse copy of the gene for the second
child

• Example:

Suppose H for Parent1 and T for Parent 2

H H T H T H
Parent 1 3 5 9 12 18 1 3 5 1
8 112 0 151 1 Offspring 1

Parent 2 4 6 8 10 15 7 9 010 1 180 7

4 6 0 Offspring 2

8
 Uniform Crossover for Integers
• Inheritance is independent of position
• Applicable for binary representation
• How to implement it (programming)
• N-point crossover ???

3/8/2024 9
 Permutation Representations
• Ordering/sequencing problems form a special type
• Task is (or can be solved by) arranging some objects
in a certain order
– Example: sort algorithm: important thing is which
elements occur before others (order)
– Example: Travelling Salesman Problem (TSP) : important
thing is which elements occur next to each other
(adjacency)
• These problems are generally expressed as a
permutation:
– if there are n variables then the representation is as a list
of n integers, each of which occurs exactly once

3/8/2024 10
 Crossover for Permutations
• Normal crossover operators will often lead to
inadmissible solutions

12345 12321
54321 54345

• Many specialised operators have been devised

which focus on combining order or adjacency
information from the two parents

3/8/2024 11
 Mutation operators for permutations

• Normal mutation operators lead to inadmissible

solutions
– e.g. bit-wise mutation : let gene i have value j
– changing to some other value k would mean that k
occurred twice and j no longer occurred
• Therefore must change at least two values
• Mutation parameter now reflects the probability
that some operator is applied once to the whole
string, rather than individually in each position

3/8/2024 12
 Insert Mutation for permutations
• Pick two allele (gene) values at random
• Move the second to follow the first, shifting
the rest along to accommodate
• Note that this preserves most of the order and
the adjacency information

3/8/2024 13
 Swap mutation for permutations
• Pick two alleles (genes) at random and swap
their positions
• Preserves most of adjacency information (4
links broken)
• disrupts order more

3/8/2024 14
 Inversion mutation for permutations
• Pick two alleles at random and then invert the
substring between them.
• Preserves most adjacency information (only
breaks two links) but disruptive of order
information

3/8/2024
 Floating-point (real values) Representation
• Many problems occur as real valued problems, e.g.
continuous parameter optimization f :  n → 

• The chromosome will be an array of floating point variables

• This can serve in optimization of a multi-variate function

y=f(x1, x2, … xm)

• Crossover is the same as in bit-string chromosomes.

• Mutation is different

3/8/2024 16
 Crossover over FP Chromosomes

x1 x2 x3 ... xm-1 xm x'1 x'2 x'3 ... x'm-1 x'm

x1 x2 x'3 ... x'm-1 x'm x'1 x'2 x3 ... xm-1 xm

3/8/2024 17
 Mutation over FP Chromosomes
• General scheme of floating point mutations
x = x1 , ..., xl → x  = x1, ..., xl
xi , xi  LBi ,UBi 

• Two kinds of FP mutations:

– Uniform
– Non-uniform

3/8/2024 18
 Uniform FP Mutation
xi drawn randomly (uniform) from LBi ,UBi 

x1 x2 … xi … xm-1 xm

Given the above chromosome X in a particular generation G:

• Each gene (variable) has a range
• Gene xi is a FP value inside chromosome X at generation G
• To mutate gene xi LB UBi
i
1. Generate random number ri1 ϵ [0, 1] xi
• ∆ = ∆L if ri1 ≤ 0.5
• ∆ = ∆U if ri1 > 0.5
• This means equal chance to go left or right ∆L ∆U
2. Generate random number ri2 ϵ [0, ∆]
• if ∆ = ∆L then xi-new = xi - ri2 ∆L = xi - LBi
• if ∆ = ∆U then xi-new = xi + ri2
∆U = UBi - xi
3/8/2024 19
 Non-uniform FP Mutation
Some methods proposed such as time-varying range of change

x1 x2 … xi … xm-1 xm

Given the above chromosome X in a particular generation G:

• Each gene (variable) has a range
• Gene xi is a FP value inside chromosome X at generation G
• To mutate gene xi
1. Generate random number ri1 ϵ [0, 1]
• y = ∆L if ri1 ≤ 0.5 LBi UBi
• y = ∆U if ri1 > 0.5
xi
2. Let ∆(t,y)
= value of mutation at generation t
= y(1-r(1-t/T)^b) ∆L ∆U
where:
• r = random number ϵ [0, 1]
• t = current generation ∆L = xi - LBi
• T = maximum number of generations
• b = dependency factor ≈ 1…5 ∆U = UBi - xi
3/8/2024 20
 Non-uniform FP Mutation
• Analysis of Equation:
∆(t,y) = value of mutation at generation t
= y(1-r(1-t/T)^b)
where:
– r = random number ϵ [0, 1]
– t = current generation
– T = maximum number of
generations
– b = dependency factor ≈ 1…5
(Controls the curve of mutation)

– At t=0:
∆(t,y) = Maximum value of mutation
– At t=T:
∆(t,y) = 0 (No mutation)

3/8/2024 21
 Crossover or Mutation?
• Decades of long debate: which one is better or
necessary?

• Answer (at least, rather wide agreement):

– it depends on the problem, but
– in general, it is good to have both
– both have different roles
– mutation-only-GA is possible, crossover-only-GA would not
work. Why??
 Mutation
• Causes movement in the search space
(local or global)
• Restores lost information to the population
• Mutation is necessary because some
important genes might be missing from all the
initial population.

3/8/2024
 Crossover
• It greatly accelerates search early in evolution
of a population
• It leads to effective combination of schemata
(subsolutions on different chromosomes)

3/8/2024
 Crossover or Mutation?
✓ Exploration: Discovering promising areas in the search space,
i.e. gaining information on the problem
✓ Exploitation: Optimising within a promising area, i.e. using
information
There is co-operation AND competition between them
• Crossover is explorative: it makes a big jump to an area
somewhere “in between” two (parent) areas
• Mutation is exploitative: it creates random small diversions,
thereby staying near (in the area of ) the parent
 Crossover OR mutation?
• Only crossover can combine information from two parents
• Only mutation can introduce new information (alleles)
• Crossover does not change the allele frequencies of the
population (thought experiment: 50% 0’s on first bit in the
population, ?% after performing n crossovers)
• To hit the optimum you often need a ‘lucky’ mutation