Performance of the Stochastic EM Algorithm for Estimating Mixture Parameters

Abstract

The aim of the present study is to evaluate the performance of the Stochastic EM (SEM) algorithm for estimating parameters in finite mixture distributions. For that purpose some simulation exercises which are based on the implementation of this algorithm are presented and they provide encouraging results. As a consequence it seems that SEM can offer a useful alternative for overcoming some well-known difficulties arising when the number of mixture components is itself an unknown parameter.