Global Asymptotic Stability Condition for Discrete-Time Complex-Valued Neural Networks

Abstract

This paper considers global stability conditions for discrete-time complex-valued recurrent neural networks, which are regarded as nonlinear dynamical feedback systems. A globally asymptotic stability condition for the networks is derived by way of a suitable choice of activation functions. The condition is shown to be successfully applied to a convex optimization problem, for which real field solution methods are generally tedious.