Neural network: Input anticipation may lead to advanced adaptation properties

Abstract

Network architecture is proposed, which is built according to principle of input anticipation. The network constantly anticipates the incoming input, compares the anticipation with the real input data and modifies its internal structure to ensure better anticipation in the future. It is argued that such network may exhibit advanced adaptation properties. © Springer-Verlag Berlin Heidelberg 2003.