Attribution Non-Commercial (BY-NC)

48 views

Attribution Non-Commercial (BY-NC)

- Crowdsourcing, Evolutionary Design, Their Combination, and Implications
- DBSR: Dynamic base station Repositioning using Genetic algorithm in wireless sensor network
- Difficult Channel Generation Using Genetic Algorithm
- efficient_energy_management_to_prolong_lifetime_of_wireless_sensor_network
- Multi-Objective Optimization Using
- 04632036
- Evaluación de sistemas renovables
- Piuleac
- GA.py
- CHAPTER 1 Summary Psihologie evolutionista
- Turtles Farewell Talk for Grad Students 7May-Revised4June2010
- genetic
- e 43062837
- A Hybrid Genetic Algorithm and Chaotic Function Model for Image Encryption
- 10.1.1.97.6042
- An Approach for Software Effort Estimation Using Fuzzy Numbers and Genetic Algorithm to Deal With Uncertainity
- Double Four-bar Crank-slider Mechanism
- 9
- Dawkins 2005..
- GA for Optimization Noiseless, Non Lubricated Helical Gear Pair

You are on page 1of 65

Genetic Algorithms:

Basic notions and some advanced topics

Francisco Herrera

Grupo de Investigación

“Soft Computing y Sistemas de Información Inteligentes”

Universidad de Granada

18071 – ESPAÑA

herrera@decsai.ugr.es

http://sci2s.ugr.es

Genetic Algorithms:

Basic notions and some advanced topics

SESSIONS

b. Advanced topics

Multimodal problems and multiple solutions

Multiobjective genetic algorithms

Memetic algorithms

Genetic Learning

Session a.

Genetic Algorithms

2. HOW TO CONSTRUCT THEM?

3. ON THE USE OF GENETIC ALGORITHMS

4. MODELS: GENERATIONAL VERSUS STEADY STATE

5. APPLICATIONS

6. EXAMPLE: TSP

7. SOFTWARE AND IMPLEMENTATIONS

8. CONCLUDING REMARKS

1. GENETIC ALGORITHMS.

INTRODUCTION

THE INGREDIENTS

What is a genetic algorithm?

Genetic algorithms

search

and learning

inspired in the process of

Natural and Genetic

Evolution

What is a genetic algorithm?

Natural Evolution

What is a genetic algorithm?

Artificial Evolution

EVOLUTIONARY COMPUTATION

based on populations whose individuals

represent solution to problems.

What is a genetic algorithm?

Artificial Evolution

There are 4 classic paradigms: John Holland

Inventor of genetic

algorithms

Professor of CS

Genetic Algorithms. 1975, Michigan University and Psychology at

the U. of Michigan.

Evolution

Strategies

Hans-Paul Schwefel Bionics & Evolutiontechnique

Universität Dortmund Technical University Berlin

http://www.bionik.tu-berlin.de/

Lawrence J. Fogel,

Natural Selection, Inc.

Inventor of Evolutionary

Programming

Stanford University.

Inventor of Genetic

Programming

F. Herrera - Introduction to Genetic Algorithms 8

Thre exist other modelos based on population evolution

What is a genetic algorithm?

Artificial Evolution

The ingredients

t reproduction t+1

selection

mutation

Crossover

(or recombination)

The evolution cycle

Selection

PARENTS

Crossover

Mutation

POPULATION

Replacement

DESCENDANTS

Genetic Algorithm Structure

Beginning (1)

t=0

Initialization P(t)

evalution P(t)

While (the stop contition is not verified) do

Beginning (2)

t=t+1

selection P’(t) from P(t-1)

P’’(t) ← crossover P’(t)

P’’’(t) ← mutation P’’(t)

P(t) ← replacement (P(t-1),P’’’(t))

evaluation P(t)

Final(2)

Final(1)

F. Herrera - Introduction to Genetic Algorithms 12

2. HOW TO CONSTRUCT A GA?

PE LEM

T

EN

Representation

ND

DE OB

Initial population

PR

Fitness function (How to evaluate a GA?)

Chromosomes selection for parents

Design of crossover operator

ON THM

TS

EN

Design of mutation operator

I

CO GOR

Chromosomes replacement

MP

Stop condition

AL

Representación

according to the evaluation and genetic the

operators.

Example: Binary representation

CHROMOSOME

GEN

F. Herrera - Introduction to Genetic Algorithms 15

Example: Binary representation

Fenotype

8 bits genotype • integer

• real number

•

• ...

• ¿Others?

Example: Real coding

real valued vector:

⎡ x1 ⎤

⎢x ⎥

X= ⎢ 2⎥

, xi ∈ R

⎢M⎥

⎢ ⎥

⎣ xn ⎦

The evaluation function associates a real

value to a vector:

f :R n → R

F. Herrera - Introduction to Genetic Algorithms 17

Example: Order representation

permutations.

Ej. Travelling salesman problem (TSP), ...

It needs special operators for obtaining a new

permutation.

Initialization

Binary string: 0 or 1 with probability 0.5

Real value: uniform on the interval

chromosomes.

Fitness function

Subrutine, simulator or other external processes (ej.

Robot experiment, ...)

It is possible to use an approximation function

(reducing the cost)

Constraint problems can introduce a penalization in

the fitness function.

dominated solutions).

HOW TO CONSTRUCT A GA?

Selection

PARENTS

Representation

Initialization

Population

Chromosomes selection

possibility for being parents.

reproduction. They can include useful genetic information in

the reproduction process.

degree of influence of the best individuals.

Strategy of selection

Tournament selection

For each parent:

Selection of the best

selective pressure and vice versa.

HOW TO CONSTRUCT A GA?

Selection

PARENTS

Representation Crossover

Initialization

Population

Crossover operator

Features:

associated to the parent features. In other case it

would be a mutation operator.

between 0.6 and 0.9, usually).

Example: Simple crosover on the

binary representation

Population: ...

recombined (example for n = 2)

cut cut

parents

1 1 1 1 1 1 1 0 0 0 0 0 0 0

descendants

1 1 1 0 0 0 0 0 0 0 1 1 1 1

Crossover operator

Example: Two points crossover

Example: uniform crossover

a b c d e f g h

a b C d E f g H

A B CDE F GH

Example: Real coding crossover

operator

Arithmetic crossover:

a b c d e f

A B CDE F

↓

(a+A)/2 (b+B)/2 (c+C)/2 (d+D)/2 (e+E)/2 (f+F)/2

Example: real coding crossover

operator BLX-α

Two chromosomes

C1 = (c11,…, c1n) y C2 = (c21,…, c2n) ,

Hk = (hk1,…, hki,…, hkn) , k =1,2

[Cmin – I·α, Cmax + I·α]

Cmin = min {c1i , c2i}

I = Cmax - Cmin , α ∈ [0,1]

Ejemplo: Operador de cruce para

representación real: BLX-α

Exploration Exploration

ai c1i c2i bi

Exploitation

Example: Crossover operator for order

representation: OX

Parent 1 Parent 2

7 3 1 8 2 4 6 5 4 3 2 8 6 7 1 5

7, 3, 4, 6, 5

Order

1 8 2 4, 3, 6, 7, 5

Hijo 1

7 5 1 8 2 4 3 6

F. Herrera - Introduction to Genetic Algorithms 33

HOW TO CONSTRUCT A GA?

Selection

PARENTS

Representation Crossover

Initialization

Population Mutation

POPULATION Fitness function

DESCENDANTS

Mutation operator

Features:

runs.

We must control the size.

It must produce valid chromosomes.

It is used with a low running probability on the descendant

obtained after the application of the crossover operator.

Example: binary mutation

before 1 1 1 1 1 1 1

after 1 1 1 0 1 1 1

Mutated gen

The mutation happens with a low running

probability per gen pm

Example: real coding mutation

•Using a gaussian/normal distribution N(0,σ),

• 0 is the mean

• σ is the typical desviation

x’i = xi + N(0,σi)

For each parameter.

Example: order representation

mutation

7 3 1 8 2 4 6 5

7 3 6 8 2 4 1 5

HOW TO CONSTRUCT A GA?

Selection

PARENTS

Representation Crossover

Initialization

Population Mutation

POPULATION Fitness function

Replacement

DESCENDANTS

Replacement strategy

Complete population

….

Stop condition

Limited CPU:

Maximum of evaluations

improvement.

Components

Selection

PARENTS

Representation Crossover

Initialization

Population Mutation

POPULATION Fitness function

Replacement

DESCENDANTS

3. MODELS: GENERATIONAL vs STEADY

STATE

complete population

parents are used with genetic operators for

getting one/two descendants.

population.

The steady state is elitist. High selection pressure

when we replace the worst individuals.

Generational model

SELECTION CROSSOVER

C1 C’1 C’’1

C2 C’2 WITH PROB. Pc C’2

… … …

CM C’M C’’M

MUTATION

WITH PROB.

Pm

H1 H1= C’’m1

REPLACEMENT

H2

… H 2 =C’m

2

With elitism (maintaining the

HM-1 …

CF.best best individual from P(t))

Herrera - Introduction to Genetic AlgorithmsH =C’’

M M 44

Steady state model

Pactual(t)

SELECTION PPAREN CROSSOVER PINTERM.

C1 TS

C’’1

C2 C’1 WITH PROB. 1

C’’2

… C’2

CM

MUTATION

WITH PROB.

Pm

t ← t+1 C1 H1= C’’1

C2 (TWO OFFSPRING

… H2=C’’m2

COMPETE WITH

H1

INDIVIDUALS IN P(t))

CM

F. Herrera - Introduction to Genetic Algorithms 45

4. APPLICATIONS

Clasification control

Learning

Structural

optimization

Trayectory

generation

Planification

1 2 m

5. EXAMPLE: TRAVELLING SALESMAN

PROBLEM

Order representation

(3 5 1 13 6 15 8 2 17 11 14 4 7 9 10 12 16)

17 cities

Objective: Sum of distance among cities.

Population: 61 chromosomes - Elitism

Crossover: OX (Pc = 0,6)

Mutation: List inversion (Pm = 0,01 – chromosome)

TSP

F. Herrera - Introduction to Genetic Algorithms 48

TSP

Optimum solution: 226.64

F. Herrera - Introduction to Genetic Algorithms 49

TSP

Optimum solution: 226,64

F. Herrera - Introduction to Genetic Algorithms 50

TSP

Iteration: 250

Iteration: 200 Costo: 231,4

Optimum

F. Herrera - Introduction to Genetic solution:

Algorithms 226,64

51

TSP

size 50 and 70 iterations

F. Herrera - Introduction to Genetic Algorithms 52

TSP

size 50 and 70 iterations

F. Herrera - Introduction to Genetic Algorithms 53

TSP

size 50 and 70 iterations

F. Herrera - Introduction to Genetic Algorithms 54

TSP

size 50 and 70 iterations

F. Herrera - Introduction to Genetic Algorithms 55

TSP

size 50 and 70 iterations

F. Herrera - Introduction to Genetic Algorithms 56

TSP

size 50 and 70 iterations

F. Herrera - Introduction to Genetic Algorithms 57

6. SOFTWARE AND IMPLEMENTATIONS

library. It contains classes for almost any kind of evolutionary computation

you might come up to at least for the ones we could think of. It is

component-based, so that if you don't find the class you need in it, it is

very easy to subclass existing abstract or concrete classes.

http://eodev.sourceforge.net/

<jjmerelo@gmail.com>

6. SOFTWARE AND IMPLEMENTATIONS

developed in the Java programming language. It provides a high-level

software environment to do any kind of Evolutionary Algorithm (EA), with

support for genetic algorithms (binary, integer and real encoding), genetic

programming (Koza style, strongly typed, and grammar based) and

evolutionary programming.

http://jclec.sourceforge.net/

for Evolutionary Computing. Soft Computing, 2007, in press.

6. SOFTWARE AND IMPLEMENTATIONS

JCLEC Features

• Wide audience. Software can be used by both expert people (EC

researchers) and novice people (students or people from another field).

• Several execution modes

• Batch mode (expert) based on a configuration file

• Interactive mode (novice) based on a GUI with charting

• Multiple individual encodings

• Linear genotype (binary, integer and real)

• Tree genotype (expression tree and syntax tree)

• Multiple crossover and mutation operators for each type of individual.

• A variety of Evolutionary Algorithms

• Classical: SGA, Steady state and CHC

• Multiobjective: SPEA2 and NSGA-II

• Niching: Clearing, sequential and fitness sharing

• Memetic: Generational and steady state

6. SOFTWARE AND IMPLEMENTATIONS

• Extensible and reusable:

• New problems can be easily added to the system.

• New components (encodings, genetic operators and/or algorithms) can

be easily added to the system.

• Multithreading

• Speeds up evaluation in multiprocessor architectures

• Parallel genetic algorithms implementation.

• Future work

• New toolboxes planned:

• Real optimization: ES, DE, MGG and GGG.

• Evolutionary neural networks.

• Native JCLEC. In complex problems, would be desirable to sacrifice

the portability of the Java version to get a bigger efficiency (some

applications increase their speed 10 times). This version has been

developed using GNU gcj (java to native) compiler.

F. Herrera - Introduction to Genetic Algorithms 61

6. SOFTWARE AND IMPLEMENTATIONS

<<interface>> <<interface>> Evolutionary system and algorithm

contains

ISystem IAlgorithm

contains

Evaluated

<<interface>> <<interface>>

IIndividual ISpecies

<<interface>>

Created

contains

ITool

Data in an Evolutionary

Individual commonalities

Algorithm

Operations performed in an EA

CONCLUDING REMARKS

Genetic Algorithms

o Based in a biological metaphor: evolution

o high applicability

o very popular

o High performance and low cost

o Powerful algorithms for a lot of applications

CONCLUDING REMARKS

EVOLUTIONARY COMPUTATION

CLASSIC PARADIGMS MODELS

ESTIMATION DISTRIBUTIION

ALGORITHMS: PBIL, EDA, ...

GENETIC EVOLUTION

PARTICLE SWARM: SOCIAL

ALGORITHMS STRATEGIES

ADAPTATION

PROGRAMMING PROGRAMMING CULTURAL EVOLUTIOARY

ALGORITHMS

DIFFERENTIAL EVOLUTION

MEMETIC ALGORITHMS

F. Herrera - Introduction to Genetic Algorithms 64

BIBLIOGRAPHY

BOOKS - Classics

D.E. Goldberg, Genetic Algorithms in Search, Optimization and Machine

Learning. Addison Wesley, 1989.

Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution

Programs. Springer Verlag, 1996.

D.B. Fogel (Ed.) Evolutionary Computation. The Fossil Record.

(Selected Readings on the History of Evolutionary Computation).

IEEE Press, 1998.

BOOK – Recent

A.E. Eiben, J.E. Smith. Introduction to Evolutionary Computation.

Springer Verlag 2003. (Natural Computing Series)

D. Whitley "A Genetic Algorithm Tutorial". Statistics and Computing, 4 (1994) 65-85.

F. Herrera, M. Lozano, J.L. Verdegay, Tackling Real-Coded Genetic Algorithms:

Operators and tools for the Behaviour Analysis. Artificial Intelligence Review 12

(1998) 265-319.

- Crowdsourcing, Evolutionary Design, Their Combination, and ImplicationsUploaded byalecperkins
- DBSR: Dynamic base station Repositioning using Genetic algorithm in wireless sensor networkUploaded byIJCSI Editor
- Difficult Channel Generation Using Genetic AlgorithmUploaded byAdam Hansen
- efficient_energy_management_to_prolong_lifetime_of_wireless_sensor_networkUploaded byRaghavendra Rakesh
- Multi-Objective Optimization UsingUploaded bysaeedullah81
- Evaluación de sistemas renovablesUploaded byJonathan Villota
- PiuleacUploaded byPiuleac1402
- Turtles Farewell Talk for Grad Students 7May-Revised4June2010Uploaded byulluaol
- 04632036Uploaded byBounty Jha
- CHAPTER 1 Summary Psihologie evolutionistaUploaded byMarius Anitoaei
- geneticUploaded byratkanasha
- A Hybrid Genetic Algorithm and Chaotic Function Model for Image EncryptionUploaded bySadique Nayeem
- GA.pyUploaded bysunilverma2010
- e 43062837Uploaded byAnonymous 7VPPkWS8O
- 10.1.1.97.6042Uploaded byesusini
- An Approach for Software Effort Estimation Using Fuzzy Numbers and Genetic Algorithm to Deal With UncertainityUploaded byCS & IT
- Double Four-bar Crank-slider MechanismUploaded byAdam Hansen
- 9Uploaded byZanariah Zainudin
- Dawkins 2005..Uploaded byPinPunk
- GA for Optimization Noiseless, Non Lubricated Helical Gear PairUploaded byshivakees
- Draft MT 461 Path Planning in Mobile RobotsUploaded byUmair Aziz
- kasarda 1Uploaded bySheikh Firdous
- e5b1ec89b298b93008d4e2f840923b5fUploaded byHector Grover G
- Combined Simulated Annealing and Genetic Algorithm to Solve Optimization ProblemsUploaded byAh M Ed
- ITjfgjhgjkhgUploaded byMaruthi Jacs
- MIKE11.pdfUploaded byTrịnh Thái Hà
- BIO30 ReportUploaded byRobelyn Murillo
- Adaptive Compliance WingUploaded byLalit Bhola
- 04 JenniferUploaded byShreya Chakraborty
- Genetic AlgorithmUploaded byShiva Thavani

- Altoids CampaignUploaded bymarkpgus
- Lecture Plan Tri 2 1415 - BAE1044 -English for Business Communication 14Oct2014Uploaded bycharlie simo
- The Wall of Willows 2Uploaded byChino Chan
- Brian Tracy Goal PlannerUploaded byScribdnera
- Blood FilmsUploaded byROHIT008RAJ
- QuestionsUploaded byMuhammad Rizam Hj Bakri
- Envisat ERS Data Access GuideUploaded bypartho_scribd
- Trump Florida Deposition - 2013Uploaded byGrant Stern
- Pomerleau v. McCarthy, CUMcv-08-408 (Cumberland Super. Ct., 2009)Uploaded byChris Buck
- U.S. Copyright Renewals, 1955 July - December by U.S. Copyright OfficeUploaded byGutenberg.org
- Bombardier i 2016Uploaded byQuaqu Ophori Asiedu
- lesson notes English year 1Uploaded byIma Khalid
- CSC Chapter 1Uploaded byPriya Srinivasan
- Hyperbaric Oxygen Therapy for Wound Healing and Limb Salvage, Robert Goldman, MDUploaded byomad pendaftaranPPDS
- 1f2040b5de902b8b323b0a6bc2f180b1 2.2 Push and Pull Sources of InnovationUploaded bybclarke113
- PRTM 2r1 WebUploaded byjiar001
- dr mamUploaded byakshay shirke
- diccionario jurídico ingles españolUploaded byLeonardo Llanos Lagos
- Topic 10Uploaded bySuhaily Mohd Ramly
- New PPT (Torture)Uploaded byMark Pesigan
- salesforce_customer_portal_implementation_guideUploaded bymanu2011
- Doubtless - Jennifer LeClaireUploaded byNOWHERE-MAN
- Sandstone Reservoir Quality Prediction the State of the Art AAPGBulletin Ajdukiewicz Lander 2010Uploaded byJody Van Rensburg
- Reelection From Curved SurfacesUploaded byShreeyansh Bhardwaj
- Thomas F. Limerick v. Carol S. Greenwald, Roslindale Cooperative Bank of Boston v. Carol S. Greenwald, 749 F.2d 97, 1st Cir. (1984)Uploaded byScribd Government Docs
- 5_6260114884906713097Uploaded bySatrio Arief
- Non Store RetailingUploaded byVanishree Pabalkar
- Lesson Plan Teaching NumbersUploaded byRandyBagoes
- Thrown Under The Bus InterviewUploaded byDavid Welch
- Final Chapter 1Uploaded byqazalbash1109588

## Much more than documents.

Discover everything Scribd has to offer, including books and audiobooks from major publishers.

Cancel anytime.