Efficient Monte Carlo sampling by parallel marginalization
Jonathan Weare, CIMS

Monte Carlo sampling methods often suffer from long correlation times. Consequently, these methods must be run for many steps to generate an independent sample.  In this talk a method is proposed to overcome this difficulty. The method utilizes information from rapidly equilibrating coarse Markov chains that sample marginal distributions of the full system.  This is accomplished through exchanges between the full chain and the auxiliary coarse chains. Results of numerical tests on the bridge sampling and filtering/smoothing problems for a stochastic differential equation are presented.  If time permits we will discuss applications of the method to a filtering problem for a bimodal model of the Kuroshio current.