Qualitative analysis of neural networks under structural perturbations

Author

Grujic, Ljubomir T. ; Michel, A.N.

Author_Institution

Dept. of Electr. & Comput. Eng., Louisiana State Univ., Baton Rouge, LA, USA

fYear

1990

fDate

1-3 May 1990

Firstpage

391

Abstract

A qualitative analysis of Hopfield neural networks with unknown random structure is developed. Simple algebraic conditions are established for structural exponential stability of x=0 of the neural network and for an estimate of its domain. By using the natural form, the quadratic form, and maximum-type form of a Lyapunov function, three differential estimates of the domain D_e of structural exponential stability of x=0 of the neural network are given. Another simple algebraic condition presented guarantees the maximum possible estimate of D_e. In all the cases bounds on the motions of the neural network in a forced regime are provided without using any information about its unknown, random structure

Keywords

learning systems; neural nets; Hopfield neural networks; Lyapunov function; differential estimates; maximum-type form; natural form; quadratic form; qualitative analysis; structural exponential stability; structural perturbations; unknown random structure; Circuit stability; Hopfield neural networks; Large-scale systems; Lyapunov method; Motion analysis; Motion estimation; Neural networks; Neurons; Stability analysis; Steady-state;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1990., IEEE International Symposium on

Conference_Location

New Orleans, LA

Type

conf

DOI

10.1109/ISCAS.1990.112054

Filename

112054