Thursday 17 Mar 2016Statistical Science Seminar: Generalising the Metropolis-Hastings Algorithm: Approaches for parallelising MCMC algorithms.

Ben Calderhead - Imperial College

H101 14:30-16:30

 Markov chain Monte Carlo methods (MCMC) are essential tools for

solving many modern-day statistical and computational problems; however, a

major limitation is the inherently sequential nature of these algorithms.

In this talk, Iıll discuss a natural generalization of the

Metropolis-Hastings algorithm that allows for parallelizing a single chain

using existing MCMC methods.  We do so by proposing multiple points in

parallel, then constructing and sampling from a finite-state Markov chain

on the proposed points such that the overall procedure has the correct

target density as its stationary distribution.  Our approach is generally

applicable and straightforward to implement.  Iıll then give an overview

of some results and directions for current and future research.

Add to calendar

Add to calendar (.ics)