Neighborhood detection using mutual information for the identification of cellular automata

Extracting the rules from spatio-temporal patterns generatedby the evolution of cellular automata (CA) usually requires a priori informationabout the observed system, but in many applications little informationwill be known about the pattern. This paper introduces a newneighborhood detection algorithm which can determine the range of theneighborhood without any knowledge of the system by introducing a criterionbased on mutual information (and an indication of over-estimation). Acoarse-to-fine identification routine is then proposed to determine the CArule from the observed pattern. Examples, including data from a real experiment,are employed to evaluate the new algorithm.

IEEE Transactions on Systems, Man and Cybernetics, Part B (Cybernetics), 36(2), 473–479. Institute of Electrical and Electronics Engineers (IEEE)