A combined gradient learning algorithm for multilayered neural networks

Author

Guozhong, Zhou ; Yaming, Sun

Author_Institution

Dept. of Electr. Eng. & Autom., Tianjin Univ., China

fYear

1991

fDate

28 Oct-1 Nov 1991

Firstpage

1492

Abstract

A combined gradient learning algorithm is developed based on the gradient and the conjugate gradient optimization algorithms. It combines the advantages of the two optimization algorithms and can greatly increase the convergence speed of learning for multilayered neural networks. It does not have a large storage requirement. The authors review the back-propagation model algorithm, the conjugate-gradient-based algorithm, and the combined gradient algorithm. Simulation results for the XOR problem and the SYMMETRY problem are presented

Keywords

conjugate gradient methods; convergence of numerical methods; learning systems; neural nets; optimisation; EXOR problem; SYMMETRY problem; XOR problem; back-propagation; combined gradient learning algorithm; conjugate gradient optimization; convergence speed; multilayered neural networks; Artificial neural networks; Automation; Control system synthesis; Convergence; Multi-layer neural network; Neural networks; Neurons; Process control; Sun; System identification;

fLanguage

English

Publisher

ieee

Conference_Titel

Industrial Electronics, Control and Instrumentation, 1991. Proceedings. IECON '91., 1991 International Conference on

Conference_Location

Kobe

Print_ISBN

0-87942-688-8

Type

conf

DOI

10.1109/IECON.1991.239120

Filename

239120