Xi'an's Og

The conference at the University of Warwick is, I think, going on quite well.There are many talks about new developments in the theory of (MC)MC. For instance, Christophe Andrieu gave a talk on particle MCMC which is a mix of sequential Monte Carlo (SMC) and MCMC. This is the opposite of population Monte Carlo in that the proposal is based on a population of particles derived by SMC and the validation is thru MCMC, with the surprising result that the moves are valid for any number of particles. (I can discuss this paper more confidently as I reviewed Roman Holenstein’s dissertation last weekend.) Several talks covered theoretical properties of MCMC algorithms of the Roberts-Rosenthal type. There also seems to be a revival of perfect sampling with an extension by Berthelsem, Roberts and Breyer of our work on mixture posteriors with Jim Hobert, about ten years ago, and a point process work by Huber and Moller that involves ghosts and shadows. (The only drag is that [public] transportation between the hotel and the university takes hours, thanks to diverse construction works that make bus schedules completely useless!)

This entry was posted on March 18, 2009 at 8:49 am and is filed under Statistics, University life. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.