Prediction of dynamical phenomena by a neural network

An adaptive information processing system capable of predicting dynamical phenomena is described. It includes a neural network-like memory, a predictor, two shift registers, and a comparator. In the memory, an internal empirical model of observed phenomena is formed. It is described by a set of memorized prototype transitions between successive states of an input time-dependent signal which can also be chaotic. System operation is demonstrated on a chaotic signal generated by the Henon map.<>