Synthesis ofBinaryCellularAutomata basedon BinaryNeuralNetworks

Thispaperstudies a simple synthesis algorithm ofdesired cellular automata(ab.CAs)basedon a binary neural network (ab.BNN).TheCAshavebinary state variable andtheBNN hasbipolar connection parameters. Inorderto realize dynamics oftheCA,thebipolar parameters ofBNN aredetermined using thegenetic algorithm. Performing basic numerical experiments weshowthat theBNNcanrealize desired dynamics withsmall numberofhidden layers foraclass ofCAs. We also consider thecaseofnoisy teacher signals. Theseresults provide basic information forapplication tosignal processing, analysis ofdigital nonlinear phenomena, andsoon.