Coevolutionary search for one-dimensional cellular automata, based on parameters related to their dynamic behaviour

Issue title: SBRN'02

Article type: Research Article

Authors: Oliveira, Gina M.B. | Asakura, Oscar K.N. | de Oliveira, Pedro P.B.

Affiliations: Pós-Graduação em Engenharia Elétrica, Universidade Presbiteriana Mackenzie, Rua da Consolação, 930 – CEP 01302-907, Consolação, São Paulo, SP, Brazil. E-mail: gina@mackenzie.br, oasakura@namy.com.br, pedrob@mackenzie.br

Abstract: The understanding of how cellular automata (CA) carry out arbitrary computations through totally local and parallel processing and how to harness their programmability is still extremely vague. In order to face this question various evolutionary methods have been used to look for cellular automata of a predefined computational behaviour. In this context, the most widely studied CA task is the density classification task (DCT), the best rule currently known for it having been obtained by a coevolutionary genetic algorithm (CGA). Extending our previous success in incorporating a parameter-based heuristic into a standard, single population genetic algorithm in the DCT, here we analyse the influence of incorporating that heuristic into the coevolutionary search. The results obtained show that the parameters can effectively help a CGA in searching for DCT rules, and suggest that the choice of the amount of bias in the search, allowed for the heuristic is more sensitive than in the standard genetic search.

Journal: Journal of Intelligent & Fuzzy Systems, vol. 13, no. 2-4, pp. 99-110, 2002/2003