A Novel Generalized Congruence Neural Networks and Its Application in Identification Simulation

Author

Yan, Tianyun ; Chen, Yong ; Jin, Fan ; Chen, Huawei

Author_Institution

Lab. of Neural Networks, Southwest Jiaotong Univ., Sichuan

fYear

0

fDate

0-0 0

Firstpage

344

Lastpage

348

Abstract

In this paper, a novel improved generalized congruence neural networks (NIGCNN) is presented, i.e. generalized piecewise derivative of error function is back propagated to adjust weights of GCNN, which solves the problem that the approximation method, adopted by GCNN and IGCNN, can not control the weight space very well. The results of approximation for sine function and identification simulation for nonlinear dynamical system show that NIGCNN is fully effective. In simulation, NIGCNN´s stability is better than that of the previous two GCNNs, and is almost the same as that of the traditional BPNN, while its convergent speed is faster than that of the traditional BPNN, and is almost the same as that of the previous two GCNNs

Keywords

approximation theory; backpropagation; neural nets; nonlinear dynamical systems; GCNN; approximation method; back propagation; error function; generalized congruence neural network; identification simulation; nonlinear dynamical system; piecewise derivative; Approximation algorithms; Approximation methods; Artificial neural networks; Convergence; Feedforward neural networks; Intelligent networks; Neural networks; Nonlinear dynamical systems; Stability; Weight control; back propagation algorithm; generalized congruence neural networks; identification simulation;

fLanguage

English

Publisher

ieee

Conference_Titel

Information, Communications and Signal Processing, 2005 Fifth International Conference on

Conference_Location

Bangkok

Print_ISBN

0-7803-9283-3

Type

conf

DOI

10.1109/ICICS.2005.1689064

Filename

1689064