Persistence and Doubling of Chaotic Attractors in Coupled 3-Cell Hopfield Neural Networks

Abstract

Two novel phenomena for unidirectionally coupled 3-cell Hopfield neural networks (HNNs) are investigated. The first one is the persistence of chaos, which means the permanency of sensitivity and infinitely many unstable periodic oscillations in the response HNN even if the networks are not synchronized in the generalized sense. Doubling of chaotic attractors is the second phenomenon realized in this study. It can be achieved when the response network possesses two stable point attractors in the absence of the driving. This feature leads to the formation of two coexisting chaotic attractors with disjoint basins. Lyapunov functions are utilized to deduce the presence of an invariant region, and the sensitivity is rigorously proved. The absence of synchronization is approved via the auxiliary system approach and analysis of conditional Lyapunov exponents. Additionally, quadruple and octuple coexisting chaotic attractors are demonstrated, and the formation of hyperchaos is discussed.