PARALLEL RECURSIVE PREDICTION ERROR ALGORITHM FOR TRAINING LAYERED NEURAL NETWORKS

被引:86
作者
CHEN, S [1 ]
COWAN, CFN [1 ]
BILLINGS, SA [1 ]
GRANT, PM [1 ]
机构
[1] UNIV SHEFFIELD,DEPT CONTROL ENGN,SHEFFIELD S1 3JD,S YORKSHIRE,ENGLAND
关键词
D O I
10.1080/00207179008934127
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Anew recursive prediction error algorithm is derived for the training of feedforward layered neural networks. The algorithm enables the weights in each neuron of the network to be updated in an efficient parallel manner and has better convergence properties than the classical back propagation algorithm. The relationship between this new parallel algorithm and other existing learning algorithms is discussed. Examples taken from the fields of communication channel equalization and nonlinear systems modelling are used to demonstrate the superior performance of the new algorithm compared with the back propagation routine. © 1990 Taylor & Francis Group, LLC.
引用
收藏
页码:1215 / 1228
页数:14
相关论文
共 12 条
[1]  
[Anonymous], 1987, LEARNING INTERNAL RE
[2]   REPRESENTATIONS OF NON-LINEAR SYSTEMS - THE NARMAX MODEL [J].
CHEN, S ;
BILLINGS, SA .
INTERNATIONAL JOURNAL OF CONTROL, 1989, 49 (03) :1013-1032
[3]   NONLINEAR-SYSTEM IDENTIFICATION USING NEURAL NETWORKS [J].
CHEN, S ;
BILLINGS, SA ;
GRANT, PM .
INTERNATIONAL JOURNAL OF CONTROL, 1990, 51 (06) :1191-1214
[4]  
CHEN S, 1989, P IEEE C ADAPTIVE AL
[5]  
Gibson G. J., 1989, P IEEE INT C AC SPEE, P1183
[6]  
Goodwin G., 1977, DYNAMIC SYSTEM IDENT
[7]   AN ADAPTIVE LEAST-SQUARES ALGORITHM FOR THE EFFICIENT TRAINING OF ARTIFICIAL NEURAL NETWORKS [J].
KOLLIAS, S ;
ANASTASSIOU, D .
IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS, 1989, 36 (08) :1092-1101
[8]   INPUT OUTPUT PARAMETRIC MODELS FOR NON-LINEAR SYSTEMS .1. DETERMINISTIC NON-LINEAR SYSTEMS [J].
LEONTARITIS, IJ ;
BILLINGS, SA .
INTERNATIONAL JOURNAL OF CONTROL, 1985, 41 (02) :303-328
[9]  
LIPPMANN R, 1987, IEEE ASSP MAGAZINE, V4
[10]  
Ljung L, 1983, THEORY PRACTICE RECU