SINGULAR-CONTINUOUS NOWHERE-DIFFERENTIABLE ATTRACTORS IN NEURAL SYSTEMS

We present a neural model for a singular-continuous nowhere-differenti able (SCND) attractors. This model shows various characteristics origi nated in attractor's nowhere-differentiability, in spite of a differen tiable dynamical system. SCND attractors are still unfamiliar in the n eural network studies and have not yet been observed in both artificia l and biological neural systems. With numerical calculations of variou s kinds of statistical quantities in artificial neural network, dynami cal characters of SCND attractors are strongly suggested to be observe d also in neural systems experiments. We also present possible informa tion processings with these attractors. (C) 1998 Elsevier Science Ltd. All rights reserved.