Discussion of stability on recurrent neural networks for nonlinear dynamic systems

Author

Lisang, Liu ; Xiafu, Peng

Author_Institution

Dept. of Electron. Inf. & Electr., Fujian Univ. of Technol., Fuzhou, China

fYear

2012

fDate

14-17 July 2012

Firstpage

142

Lastpage

145

Abstract

Stability analysis is a most important problem in the dynamic analysis of dynamical systems. The stability properties and dynamic behavior of the recurrent neural network for nonlinear dynamic system modeling directly determine its engineering applications. In this paper, based on Lyapunov stability theory, the stability problems of recurrent neural networks (RNN) and its general stability conditions are discussed. And a novel diagonal recurrent neural network with output feedback (O-DRNN) is proposed as an concrete example, analyzing its stability as well as the range of learning rate.

Keywords

Lyapunov methods; feedback; nonlinear dynamical systems; recurrent neural nets; Lyapunov stability theory; O-DRNN; RNN; diagonal recurrent neural network-with-output feedback; dynamical system dynamic analysis; engineering applications; learning rate; nonlinear dynamic system modeling; stability analysis; stability conditions; Learning systems; Lyapunov methods; Mathematical model; Nonlinear dynamical systems; Recurrent neural networks; Stability criteria; diagonal recurrent neural network; nonlinear dynamic system; stability;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer Science & Education (ICCSE), 2012 7th International Conference on

Conference_Location

Melbourne, VIC

Print_ISBN

978-1-4673-0241-8

Type

conf

DOI

10.1109/ICCSE.2012.6295045

Filename

6295045