|
Activity Number:
|
491
|
|
Type:
|
Invited
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
ENAR
|
| Abstract - #305024 |
|
Title:
|
Split-Merge Markov Chain Monte Carlo for a Nonconjugate Dirichlet Process Mixture Model
|
|
Author(s):
|
Sonia Jain*+ and Radford Neal
|
|
Companies:
|
University of California, San Diego and University of Toronto
|
|
Address:
|
Division of Biostatistics and Bioinformatics, La Jolla, CA, 92093-0717,
|
|
Keywords:
|
Bayesian mixture model ; Markov chain Monte Carlo ; split-merge moves ; nonconjugate prior
|
|
Abstract:
|
A major impediment in designing Markov chain Monte Carlo algorithms for nonconjugate models is the computational difficulty that arises when the model is no longer analytically tractable. We propose a new nonincremental Markov chain sampling technique that efficiently clusters heterogeneous data by splitting and merging mixture components of a nonconjugate Dirichlet Process mixture model. Our method, which is a generalization of our conjugate, split-merge, Metropolis-Hastings procedure, will accommodate models with a specific type of nonconjugate prior---the conditionally conjugate family of priors. Appropriate Metropolis-Hastings split-merge proposal distributions are obtained by utilizing properties of a restricted Gibbs sampling scan. Highlights from a simulation study will be shown.
|
