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

DOI10.1142/S0219876218500123zbMATH Open1404.65008arXiv1506.07450OpenAlexW2964185649WikidataQ60656248 ScholiaQ60656248MaRDI QIDQ4563129FDOQ4563129

Monika Pietrowska, Michał Marczyk, Joanna Polanska, Andrzej Polański, Piotr Widlak

Publication date: 6 June 2018

Published in: International Journal of Computational Methods (Search for Journal in Brave)

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.

Full work available at URL: https://arxiv.org/abs/1506.07450

Recommendations

• Initializing the EM algorithm in Gaussian mixture models with an unknown number of components
• EM for mixtures
• Choosing starting values for the EM algorithm for getting the highest likelihood in multivariate Gaussian mixture models
• A Gaussian mixture model based \(k\)-means to initialize the EM algorithm
• Choosing initial values for the EM algorithm for finite mixtures

zbMATH Keywords

EM algorithmdynamic programmingGaussian mixturesmass spectra

Mathematics Subject Classification ID

Point estimation (62F10) Dynamic programming (90C39)

Cites Work

• MCLUST: Software for model-based cluster analysis
• Estimating the dimension of a model
• Finite mixture models
• Computational aspects of fitting mixture models via the expectation-maximization algorithm
• Title not available (Why is that?)
• Choosing starting values for the EM algorithm for getting the highest likelihood in multivariate Gaussian mixture models
• Choosing initial values for the EM algorithm for finite mixtures
• Model-based cluster and discriminant analysis with the MIXMOD software
• Consistency of the Maximum Likelihood Estimator in the Presence of Infinitely Many Incidental Parameters
• A likelihood-based constrained algorithm for multivariate normal mixture models
• On the approximation of curves by line segments using dynamic programming
• A profile likelihood method for normal mixture with unequal variance
• Initializing the EM algorithm in Gaussian mixture models with an unknown number of components
• Measuring the component overlapping in the Gaussian mixture model
• On the Bumpy Road to the Dominant Mode
• Least Squares Estimators of Peptide Species Concentrations Based on Gaussian Mixture Decompositions of Protein Mass Spectra

Cited In (3)

• Method for mass spectrometry spectrum deisotoping based on fuzzy inference systems
• An effective strategy for initializing the EM algorithm in finite mixture models
• Randomly initialized EM algorithm for two-component Gaussian mixture achieves near optimality in \(O(\sqrt{n})\) iterations

Uses Software

• mclust
• Mixmod
• ElemStatLearn
• JAMM

This page was built for publication: Initializing the EM algorithm for univariate Gaussian, multi-component, heteroscedastic mixture models by dynamic programming partitions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4563129)