Evolutionary Design of Rule-Changing Cellular Automata guided by Parameter indicating Propagation of Information

Abstract

A method for designing the transition rules of cellular automata using genetic algorithms is described. Rule-changing cellular automata are expected to perform density classification tasks more effectively than ordinary cellular automata. We propose a method for designing high performance rule-changing cellular automata. This method uses a new parameter that indicates the propagation of information. Experimental results for density classification tasks show that the proposed method performs better than the previous method.