 |
 |
IMPERIAL COLLEGE STATISTICS SECTION SEMINARS This week's seminar will be held on Friday 3rd December at 2pm in Room 140 of the Huxley Building on the South Kensington campus. All are welcome. Marginal Maximum A Posteriori Parameter Estimation using MCMC Arnaud Doucet (Cambridge University) Markov chain Monte Carlo (MCMC) methods, while facilitating the solution of many complex problems in Bayesian inference, are not currently well adapted to the problem of marginal maximum a posteriori (MMAP) estimation -the estimation of the maximum a posteriori parameters of a marginal distribution-, especially when the number of parameters is large. Indeed, it usually involves solving a complex joint integration/optimization problem. We present here a simple and novel Markov Chain Monte Carlo (MCMC) strategy, called State-Augmentation for Marginal Estimation (SAME), which leads to MMAP estimates for Bayesian models. We illustrate the simplicity and utility of the approach for missing data problems including Hidden Markov models and blind deconvolution of impulsive processes. Some global convergence results are presented. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%