Development of Recurrent Perceptron Learning Algorithm for Second-Order Cellular Neural Networks

Abstract

This paper develops a method to estimate the set of weights for SOCNNs using Recurrent Perceptron Learning Algorithm. By integrating not only the First-order input and output signals but also the Second-order input and output signals into a general input signal, the research team transformed the networks of SOCNNs into an equivalent structure with the traditional Perceptron Networks. From there, the parameters of SOCNNs can be determined by the supervised learning method. The paper also simulates SOCNNs on MATLAB to check the correctness and efficiency of the proposed algorithm.