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