SOME APPLICATIONS OF STOCHASTIC APPROXIMAION METHOD

Abstract

Stochastic approximation method first introduced by Robbins-Monro [10] has been proved to be very useful for a learning system in the sense that its algorithm is very simple and that, at any given time in the learning process, the past samples are not required to retain in memory as shown by Albert and Gardner [1] , Blum [3] etc. In this paper we are concerned with its applications . After giving some preliminaries and notations in Section 2, we shall treat in Section 3 the problem of finding threshold elements, which is fundamental in pattern classification . In Section 4 we shall consider the problem of parameter identification in linear system with an additive noise.