Crossover in grammatical evolution: A smooth operator?

Michael O’Neill, Conor Ryan

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Grammatical Evolution is an evolutionary algorithm which can produce code in any language, requiring as inputs a BNF grammar definition describing the output language, and the fitness function. The usefulness of crossover in GP systems has been hotly debated for some time, and this debate has also arisen with respect to Grammatical Evolution. This paper serves to analyse the crossover operator in our algorithm by comparing the performance of a variety of crossover operators. Results show that the standard one point crossover employed by Grammatical Evolution is not as destructive as it might originally appear, and is useful in performing a global search over the course of entire runs. This is attributed to the fact that prior to the crossover event the parent chromosomes undergo alignment which facilitates the swapping of blocks which are more likely to be in context.

Original languageEnglish
Title of host publicationGenetic Programming - European Conference, EuroGP 2000 Edinburgh, Scotland, UK, April 15-16, 2000 Proceedings
EditorsWilliam B. Langdon, Peter Nordin, Riccardo Poli, Julian Miller, Wolfgang Banzhaf, Terence C. Fogarty
PublisherSpringer Verlag
Pages149-162
Number of pages14
ISBN (Print)3540673393
DOIs
Publication statusPublished - 2000
Event3rd European Conference on Genetic Programming, EuroGP 2000 - Edinburgh, United Kingdom
Duration: 15 Apr 200016 Apr 2000

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume1802
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd European Conference on Genetic Programming, EuroGP 2000
Country/TerritoryUnited Kingdom
CityEdinburgh
Period15/04/0016/04/00

Fingerprint

Dive into the research topics of 'Crossover in grammatical evolution: A smooth operator?'. Together they form a unique fingerprint.

Cite this