Talker: Arnaud Doucet (Oxford U.)
Title: Efficient Implementation of MCMC When Using An Unbiased
Likelihood Estimator
Abstract: Particle MCMC methods is a class of MCMC methods relying
on high dimensional proposals built using particle methods. In
the first part of the talk I will review the two main particle MCMC
schemes that have been proposed in the literature and discuss some
recent developments on how one to tune optimally these algorithms.
In the second part of the talk I will show how one can harness the
remarkable theoretical properties of particle MCMC methods to obtain
a perfect simulation procedure.
- slides
- video