An approach to prediction of spatio-temporal patterns based on binary neural networks and cellular automata

This paper studies application of binary neural networks (BNN) to prediction for spatio-temporal patterns. In the approach, we assume that the objective spatio-temporal patterns can be approximated by a cellular automaton (CA). Teacher signals are extracted from a part of objective pattern and are used for learning of the BNN. The BNN is used to govern dynamics of CA that outputs prediction patterns. Performing basic numerical experiments, we have investigated relation among the number of teacher signals, the number of hidden neurons and prediction performance. The results provide basic information for development of robust prediction method for digital spatio-temporal patterns.