Searching for complex CA rules with GAs

Abstract

AbstractAn important problem in the theory of cellular automata (CA) is the research of complex rules, the parameterization of the rule space, and the description of the CA dynamic behavior. In this article, the authors used evolutionary techniques in the search of CA with specific characteristics, in order to understand how their data are related to the previous findings in this sector. Using as fitness function, the variance of the input‐entropy, a genetic algorithm was built to search complex rules for multistate CA. The algorithm used is efficient and many complex rules have been found. Many of these rules present interesting characteristics. Gliders with large period and large spatial dimensions may be observed. The experiments performed show that complex rules have several λ values, which do not reside in a sharply defined region; on the contrary, these values are extended on very large regions of the λ parameter. The results provide also a view of the CA rule space. © 2003 Wiley Periodicals, Inc.