Crossover in grammatical evolution: The search continues

Michael O'Neill, Conor Ryan, Maarten Keijzer, Mike Cattolico

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

Abstract

Grammatical Evolution is an evolutionary automatic programming algorithm that can produce code in any language, requiring as inputs a BNF grammar definition describing the output language, and the fitness function. The utility 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 continue an analysis of the crossover operator in Grammatical Evolution by looking at the result of turning off crossover, and by exchanging randomly generated blocks in a headless chicken-like crossover. Results show that crossover in Grammatical Evolution is essential on the problem domains examined. The mechanism of one-point crossover in Grammatical Evolution is discussed, resulting in the discovery of some interesting properties that could yield an insight into the operator's success.

Original languageEnglish
Title of host publicationGenetic Programming - 4th European Conference, EuroGP 2001, Proceedings
EditorsJulian Miller, Marco Tomassini, Pier Luca Lanzi, Conor Ryan, Andrea G.B. Tettamanzi, William B. Langdon
PublisherSpringer Verlag
Pages337-347
Number of pages11
ISBN (Electronic)3540418997, 9783540418993
DOIs
Publication statusPublished - 2001
Event4th European Conference on Genetic Programming, EuroGP 2001 - Lake Como, Italy
Duration: 18 Apr 200120 Apr 2001

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume2038
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference4th European Conference on Genetic Programming, EuroGP 2001
Country/TerritoryItaly
CityLake Como
Period18/04/0120/04/01

Fingerprint

Dive into the research topics of 'Crossover in grammatical evolution: The search continues'. Together they form a unique fingerprint.

Cite this