Stability analysis of recurrent neural networks with time-varying activation functions

The dynamical behavior of a single layer recurrent neural network without hidden neurons has been investigated intensively and its stability has been analyzed using the Lyapunov method. Since the pioneering work of Hopfield many modified versions of the original Hopfield network have been suggested and their stability has been proven. In this paper we generalize these results to the case of a time-varying activation function, which is very useful in the field of parameter estimation and communications.