A1210
Title: Generalized Markov chain importance sampling methods
Authors: Quan Zhou - Texas A&M University (United States) [presenting]
Abstract: First, it is shown that for any multiple-try Metropolis algorithm, one can always accept the proposal and evaluate the important weight that is needed to correct for the bias without extra computational cost. This results in a general, convenient, and rejection-free Markov chain Monte Carlo (MCMC) scheme that extends beyond the conventional Metropolis-Hastings framework. Second, by further leveraging the importance sampling perspective on Metropolis-Hastings algorithms, an alternative importance sampling-based MCMC sampler is proposed on discrete spaces, along with a general theory on its complexity. Numerical examples suggest that the proposed algorithms are consistently more efficient than the original Metropolis-Hastings versions.
