Fall 2009 Thursday, October 15   Speaker: Yves Atchade Ph.D. The University of Michigan Title: "Adaptive MCMC: How Efficient Is It?" Abstract: Markov Chain Monte Carlo (MCMC) is a popular technique to construct Markov chains for Monte Carlo simulation purposes. The technique can sometimes be improved upon by optimizing on the fly the transition kernels involved in the algorithms, an approach known as adaptive MCMC. The discussion in this talk will focus around the central limit theorem and the estimation of asymptotic variance for adaptive MCMC. We will see that carefully designed, adaptive MCMC samplers behave similarly as their nonadaptive counterpart. Practical aspects of the algorithms will be discussed as well.

Copyright c. 2009 The University of Iowa. All Rights Reserved. Photography c. UI Photo Service and University Relations, University of Iowa This page was last modified on Thursday, 08-Oct-2009 09:16:24 CDT . /me