Dependent feature trees for density approximation II. Maximum likelihood clustering

Abstract

Abstract In this paper, maximum likelihood clustering for the decomposition of mixture density of data into its normal component densities is considered. Optimal dependent feature trees for approximating the densities can be constructed using criteria of mutual information and distance measures. By defining different types of nodes in a general dependent feature tree, maximum likelihood equations are developed for the estimation of parameters of mixture density using fixed-point iterations. Furthermore, the field structure of the data is also taken into account in developing maximum likelihood equations.