|
Activity Number:
|
135
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 7, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #306222 |
|
Title:
|
Recent Developments in Population Monte Carlo
|
|
Author(s):
|
David Stephens*+
|
|
Companies:
|
Imperial College London
|
|
Address:
|
Department of Mathematics, Huxley Building, London, SW7 2AZ, UK
|
|
Keywords:
|
Markov chain Monte Carlo ; sequential Monte Carlo ; population algorithms ; mixture models ; gene expression profiles
|
|
Abstract:
|
We review population approaches to Markov chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC) with particular application to mixture modeling and cluster analysis. The Bayesian inference problem for mixtures when the data being modeled are large in number and potentially high-dimensional is especially challenging, but this is precisely the context that coincides with the analysis of gene expression profiles. Assuming a regression model for the underlying profiles, we demonstrate that population MC methods provide a computationally feasible method of solution for the resulting mixtures of regressions problem. We illustrate the implementation of population MC methods on simulated data and several real datasets from microarray experiments investigating the functional genomics of different organisms.
|
