Parsimonious Gaussian mixture models of diagonal family for binned data clustering: Mixture approach

Abstract

Binning of data in cluster analysis has advantages both in deducing the computation cost and taking into account the localization imprecision of data. In cluster analysis, basing on Gaussian mixture models is a powerful approach, among which two most common model-based cluster approaches are mixture approach and classification approach. Mixture approach estimates the model parameters by maximizing the likelihood by EM algorithm. According to eigenvalue decomposition of the variance matrices of the mixture components, parsimonious Gaussian mixture models can be generated. Choosing a proper parsimonious model can provide good result with less computation time. In this paper, we present EM algorithms applied to binned data in diagonal parsimonious models case.

10 Figures and Tables

Cite this paper

@article{Wu2011ParsimoniousGM, title={Parsimonious Gaussian mixture models of diagonal family for binned data clustering: Mixture approach}, author={Jingwen Wu and Hani Hamdan}, journal={2011 IEEE 12th International Symposium on Computational Intelligence and Informatics (CINTI)}, year={2011}, pages={385-390} }