A dynamic model for neural networks

Abstract

We present a novel dynamical model for neural networks. The model consists of two sets of differential equations: one for the hidden layers and the other for the output layer. Since the model parameters (or weights) are assumed unknown, one may choose these parameters so that the neural output behaves in a certain prescribed manner. The proposed model can be very useful for many engineering applications such as communication, control and system identification. Numerical examples are given to show the effectiveness of the proposed neural model.