Optimization of learning algorithms for Chaotic Diagonal Recurrent Neural Networks

Abstract

The traditional solutions of weight training were various derivation method in Chaotic Diagonal Recurrent Neural Networks model and its momentum gradient learning algorithm. But its deduced the precise of all the weight, without the discrete moment k. In this paper, an optimization design of sampling time k was carried out the derivation of the weight training, and a revised mathematical model was used. Simulation and results demonstrated that the optimization of sampling time k could increase the prediction accuracy and the method had generalizations in other prediction.