Initializing EM algorithm for univariate Gaussian, multi-component, heteroscedastic mixture models by dynamic programming partitions

2015/06/24 by Andrzej Polanski, Polanski, Andrzej, Michal Marczyk +7
Mathematics · #Applications (stat.AP) #FOS: Computer and information sciences #I.2.8 #I.6.5 #stat.AP

paper · pdf · doi:10.48550/arxiv.1506.07450

21 pages, 2 figures

arxiv created 2015/08/02 · arxiv updated 2015/08/04

Abstract

Setting initial values of parameters of mixture distributions estimated by using the EM recursive algorithm is very important to the overall quality of estimation. None of the existing methods is suitable for mixtures with large number of components. We present a relevant methodology of estimating initial values of parameters of univariate, heteroscedastic Gaussian mixtures, on the basis of the dynamic programming algorithm for partitioning the range of observations into bins. We evaluate variants of dynamic programming method corresponding to different scoring functions for partitioning. For simulated and real datasets we demonstrate superior efficiency of the proposed method compared to existing techniques.

Related

• Randomly initialized EM algorithm for two-component Gaussian mixture achieves near optimality in O(√(n)) iterations
• Convergence and Optimality of the EM Algorithm Under Multi-Component Gaussian Mixture Models
• Bayesian estimation of dynamic weights in Gaussian mixture models
• A simple multithreaded implementation of the EM algorithm for mixture models
• Model Selection for Gaussian Mixture Models