An Approach to Searching for Two-Dimensional Cellular Automata for Recognition of Handwritten Digits

摘要: One of the contexts in which cellular automata have clearly demonstrated their effectiveness has been problems involving strong and explicit spatial constraints, as happens pattern formation growth. By analogy, attempts to use recognition also used literature some progress made. However, general, they still represent more an unfulfilled promise, due lack a model would naturally fit in, effective ways implement it, generality available approaches. Here, experimental results are reported direction using task handwritten digit recognition, evolutionary algorithm searches for two-dimensional rules that transform given image into match, close possible, prototype family, so that, closer better input image. Although might fall shorter than consolidated commercial techniques task, approach presented is quite attractive terms efficacy level it allowed achieve, because its simplicity, suggests potential from perspective other domains.