##Statistics Seminar##\\ Department of Mathematical Sciences
^ **DATE:**|Thursday, October 25, 2018 |
^ **TIME:**|1:15pm -- 2:15pm |
^ **LOCATION:**|WH 100E |
^ **SPEAKER:**|Fang Yuan, Binghamton University |
^ **TITLE:**|Dirichlet Process Mixtures of Multivariate Normal-Inverse Gaussian Distributions |
\\
**Abstract**
An expectation-maximization framework for clustering using
finite mixture models can sometimes yield uncertainty in deciding the
number of clusters. A Dirichlet process mixture model can alleviate
this difficulty of finding the correct number of mixture components by
inferring the number of clusters directly in a Bayesian framework. In
this talk, I will discuss the Dirichlet process as well as the general
framework for Dirichlet process mixture models. Implementation of a
Dirichlet process mixture of Gaussian distributions will be presented
and the generalization of this to a Dirichlet process mixture of
Multivariate Normal Inverse Gaussian (MNIG) distribution will be
discussed in detail. An algorithm for clustering skewed data based on
a Dirichlet process mixture of MNIG distributions will be discussed.