Parallel and interacting Markov chains Monte Carlo method (original) (raw)
Abstract: In many situations it is important to be able to propose NNN independent realizations of a given distribution law. We propose a strategy for making NNN parallel Monte Carlo Markov Chains (MCMC) interact in order to get an approximation of an independent NNN-sample of a given target law. In this method each individual chain proposes candidates for all other chains. We prove that the set of interacting chains is itself a MCMC method for the product of NNN target measures. Compared to independent parallel chains this method is more time consuming, but we show through concrete examples that it possesses many advantages: it can speed up convergence toward the target law as well as handle the multi-modal case.
Submission history
From: Fabien Campillo [view email] [via CCSD proxy]
[v1] Thu, 5 Oct 2006 12:35:29 UTC (298 KB)