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

TL;DR

This paper introduces a dynamic programming-based method for initializing parameters in univariate heteroscedastic Gaussian mixture models, improving estimation quality especially for models with many components.

Contribution

It proposes a novel initialization technique using dynamic programming for univariate heteroscedastic Gaussian mixtures, outperforming existing methods in efficiency.

Findings

Superior efficiency on simulated datasets

Effective for large component mixtures

Improves EM algorithm initialization quality

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.