Finite mixture models and model-based clustering

Abstract

Abstract

Kyoung-Min Kim The Department of Statistics The Graduate School Seoul National University

Finite mixture models aims at identifying clusters of individuals who show

similar patterns. The method is having been used in a variety of fields,

especially in medicine to explain the idea of heterogeneity of treatment

effects on population. The number of mixture components is typically not

known and has to be chosen.

To solve this problem, EM algorithm-based approaches is

considered. We will review details of mixture models and model-based

clustering. Furthermore, we will provide an overview of several challenges

that have been only partially resolved.

Note: Writing this paper, I mainly refer to "Finite mixture models and model-based clustering"(Melnykov, 2010)

Keyword : EM algorithm, model selection, variable selection, diagnostics, two-dimensional gel electrophoresis data, magnitude magnetic resonance images.

Student Number : 2014-20298