You are on page 1of 17

1 of 17

CHAPTER 1
EVOLUTIONARY ALGORITHMS
Index

Introduction Introduction to
Classification Evolutionary Algorithms
A Basic EA

Nomenclature

Customization

Example

Advantages and
Drawbacks
University of Málaga, Spain
Bibliography
Enrique Alba
2 of 17

CHAPTER 1
Introduction
Index
 Metaheuristic algorithms for optimization,
Introduction
search and machine learning
Classification

A Basic EA
 Techniques inspired in natural processes
including survival of the fittest and
Nomenclature
recombination of traits
Customization

Example  Mathematically, EA’s are numerical methods


Advantages and
Drawbacks  Computationally, EA’s are software programs
Bibliography
 For applications, EA’s are just tools to be used
3 of 17

CHAPTER 1
Introduction
Index
 Objective of a global optimization problem:
Introduction  
f ( x )  max : find a vector x *
  
Classification such that x  M : f ( x )  f ( x*) : f *
A Basic EA

Nomenclature

Customization

Example

Advantages and
Drawbacks

Bibliography
 Minimizing is also possible
 Vectors can map to other data structures
4 of 17

CHAPTER 1
High Level Thoughts
Index
High degree of applicability of evolutionary
Introduction algorithms to many complex tasks:
 Numeric Optimization
Classification
 Graph Partitioning
A Basic EA
 VLSI Circuits Design
Nomenclature  Desing of Neural Networks
Customization  Telecommunications
Example  Problems in Engineering
...
Advantages and
Drawbacks  Problem: Too much focus on the application,
the implementation and low efficiency
Bibliography
 Solution: Common design and analysis of new
hybrid and parallel algorithms
5 of 17

CHAPTER 1 Where Optimization Problems


can be Found?
Index

Introduction

Classification

A Basic EA

Nomenclature

Customization

Example

Advantages and
Drawbacks

Bibliography
6 of 17

CHAPTER 1
Search Techniques and EA’s
Index

Introduction Optimization Algorithms

Classification
Exact Metaheuristic
A Basic EA

Nomenclature
Calculus Enumerative Trajectory Population
Customization

Example Direct Indirect DP SA EA


B&B VNS ACO
Advantages and Newton
Drawbacks TS
Greedy
Bibliography
7 of 17

CHAPTER 1 Soft Computing


(Computational Intelligence)
Index

Introduction

Classification
Computational GA
A Basic EA Intelligence
EP
Nomenclature
Evolutionary Computation ES
Customization

Example Artificial Neural Networks GP

Advantages and Fuzzy Logic CS


Drawbacks

Bibliography
8 of 17

CHAPTER 1 Basic Working Principles in an


Evolutionary Algorithm (I)
Index

Introduction Evolutionary Algorithm


Classification t := 0;
A Basic EA initialize [P(t)];
evaluate [P(t)];
Nomenclature
while not end condition do
Customization P’(t) := variation [P(t)];
Example evaluate [P’(t)];
P(t+1) := selection [P’(t)  Q];
Advantages and
Drawbacks t := t +1 ;
end while
Bibliography
9 of 17

CHAPTER 1 Basic Working Principles in an


Evolutionary Algorithm (II)
Index  Panmictic (centralized)
 Structured (decentralized)
Introduction
EVOLUTIONARY ALGORITHM
t := 0;
P(0) : a1 (0),..., a  (0) G  ;
Classification  
initialize:
P (0) :  a1 (0) ,..., a  (0) ;
A Basic EA  
evaluate:
Nomenclature while no  P (t )  do
selection: P (t ) : s  s P (t )  ;
Customization
variation ops.: P (t ) :  v P (t )  ;
 
P (t ) :  a1(0) ,..., a (0) ;
Example
evaluate:
Advantages and replacement: P(t  1) : r r P (t )  Q  ;
Drawbacks
t := t +1 ;
Bibliography end while
 Generational (genGA)
 Steady State (ssGA)
10 of 17

CHAPTER 1 Nomenclature and


Details of one Individual
Index
Genotype The code, devised to represent the parameters of the problem in the form of a string
Chromosome One encoded string of parameters (binary, Gray, floating point numbers, etc...)
Introduction Individual One or more chromosomes with an associated fitness value
Gene The encoded version of a parameter of the problem being solved
Classification Allele Value which a gene can assume (binary, integer, real, or even complex data structures)
Locus The position that the gene occupies in the chromosome
Phenotype Problem version of the genotype (algorithm version), suited for being evaluated
A Basic EA
Fitness Real value indicating the quality of an individual as a solution to the problem
Environment The problem. This is represented as a function indicating the suitability of phenotypes
Nomenclature Population A set of individuals with their associated statistics (fitness average, Hamming dist...)
Selection Policy for selecting one individual from the population (selection of the fittest, ...)
Customization Crossover Operation that merges the genotypes of two selected parents to yield two new children
Mutation Operation that spontaneously changes one or more alleles of the genotype
Example Individual
Chromosome CHROMOSOME FITNESS
Advantages and
Drawbacks
Genes Gene 8 Gene 7 Gene 6 Gene 5 Gene 4 Gene 3 Gene 2 Gene 1 Gene 0

Bibliography
Alelles 1 1 0 1 0 1 0 0 1 0 1 0 1 1 1 1 0 0 0 1 1 0 0 0 1 0 1
Loci 2 1 0 2 1 0 2 1 0 2 1 0 2 1 0 2 1 0 2 1 0 2 1 0 2 1 0
11 of 17

CHAPTER 1
Application Methodology
Index
In order to use one EA several steps of instantiation are needed:
Introduction
 Definition of the genotype (encoding the problem variables)
Classification
(binary, real, syntactic tree, constant/variable length, etc.)
A Basic EA
 Designing an appropriate fitness function
Nomenclature
(including problem knowledge)
Customization

Example  Definition of the variation operators and their parameters

Advantages and (selection, recombination, mutation, replacement, etc.)


Drawbacks
 Definition of the stopping criterion
Bibliography
(max number of iterations, larger acceptable error, etc.)
12 of 17

CHAPTER 1
Illustrating the Behavior of a GA
Index

Introduction SEQUENTIAL GENETIC ALGORITHM


t := 0;
initialize : P(0) : a1 (0),..., a  (0) G  ;
Classification
 

evaluate : P (0) :  a1 (0) ,..., a  (0) ;


 
A Basic EA
while no  P (t )  do
Nomenclature
select : P (t ) : s  s P (t )  ;
Customization recombine : P (t ) :  c P (t )  ;
Example mutate : P (t ) : m m P (t )  ;
 
evaluate : P (t ) :  a1(0) ,..., a(0)  ;
Advantages and
Drawbacks replace : P (t  1) : r r P (t )  Q  ;
t := t +1 ;
Bibliography
end while
13 of 17

CHAPTER 1 A GA Example:
Maximize f(x)=x2 (I)
Index
String# String Fitness % of the Total
Introduction 1 01101 169 14.4
2 11000 576 49.2
3 01000 64 5.5
Classification 4 10011 361 30.9
Total 1170 100.0

A Basic EA
1. Roulette Wheel Selection (RW) 2. Single Point Crossover (SPX)
fi pc  [0.6 .. 1.0]
Nomenclature PSi = n
 fj
parents offspring
j=1 01|101 (169) 01000 (64)
Customization 10011 11|000 (576) 11101 (841)
31% 01101
01000
3. Mutation
Example 5% 14%
pm  [0.001 .. 0.1]
st
Mutation of the 1 allele
Advantages and
Drawbacks 11000
0 1 0 0 0 1 1 0 0 0
50% 4 3 2 1 0 4 3 2 1 0

fitness=64  fitness=576
Bibliography
14 of 17

CHAPTER 1 A GA Example:
Maximize f(x)=x2 (II)
Index
str# genotype phen. fitness pr. sel.
Introduction 1 11001001 201 40,401 0.28 P(i)
2 11111110 254 64,516 0.45
Classification 3 00000001 1 1 7e10-6 avg. fitness=35,735.75
4 11000011 195 38,025 0.26
A Basic EA
Select after selection Cross after crossover Mutat. after mutation
Nomenclature
111111-10 111111-11 11111111
11111110 11111110 11111110
Customization
11001001 (1)1001001 (0)1001001
110000-11 pc=0.5 110000-10 pm=1/32 11000010
Example
str# genotype phen. fitness pr. sel.
Advantages and
Drawbacks 1 11111111 255 65,025 0.37 P(i+1)
2 11111110 254 64,516 0.37
Bibliography 3 01001001 73 5,329 0.03 avg. fitness=43,126.50
4 11000010 194 37,636 0.22
15 of 17

CHAPTER 1 EA’s
Advantages and Drawbacks
Index
 Common advantages of all EA’s:
Introduction  They work on a problem codification
Classification  Very high applicability
A Basic EA  They work out mutliple solutions
Nomenclature  Multipoint-like search (population)
Customization  Drawbacks of EA’s:
Example  Don’t ensure locating the optimum
Advantages and  Many open research lines in the field:
Drawbacks
- Problems in which they are
Bibliography
better than other algorithms
- Theoretical background
16 of 17

CHAPTER 1 PEA’s;
Just a Glimpse
Index
 Complex applications: sequential EA’s have
Introduction high computational demands
Classification  Techniques to improve the efficiency:
A Basic EA
 New Operators
 Improved Representations
Nomenclature
 Tuning of the Fitness Function
Customization  Using Decentralized Models
Example
 Using Physical Parallelism
...
Advantages and Decentralized EA’s: a reduction in the
Drawbacks
needed effort is usually achieved
Bibliography
 Parallel EA’s: make more search steps per
time unit
17 of 17

CHAPTER 1
Bibliography
Index Goldberg D.E. (1989).
Genetic Algorithms in Search, Optimization and Machine Learning.
Introduction Addison-Wesley.
Bäck T., Schwefel H.P. (1993).
Classification
“An Overview of Evolutionary Algorithms for Parameter Optimization”.
Evolutionary Computation, 1 (1), 1-23.
A Basic EA
Bäck T. (1996).
Nomenclature Evolutionary Algorithms in Theory and Practice.
Oxford University Press.
Customization
Michalewicz Z.(1998, third edition).
Genetic Algorithms+Data Structures=Evolution Programs.
Example
Springer-Verlag.
Advantages and Bäck T., Hammel U., Schwefel H.P. (1997).
Drawbacks “Evolutionary Computation: Comments on the History and Currrent State”.
IEEE Transactions on Evolutionary Computation 1(1) 3-17.
Bibliography
Blum C., Roli A. (2003).
“Metaheuristics in CO: Overview and Conceptual Comparison”.
ACM Computing Surveys, 35(3), 268-308.

You might also like