Proximal Markov Chain Monte Carlo: Convex Optimisation Meets Stochastic Sampling

Convex optimisation and stochastic sampling are two powerful methodologies for performing statistical inference in inverse problems related to signal and image processing. It is widely acknowledged that these methodologies can complement each other very well; yet they are generally studied and used separately. In this talk I will discuss the potential for synergy between them and show some examples of how they can be combined to produce powerful Bayesian inference algorithms.

Marcelo Pereyra

Marcelo Pereyra was born in Buenos Aires, Argentina. He studied electrical engineering received a double MEng degree from ITBA (Argentina) and INSA Toulouse (France), together with an MSc degree from INSA Toulouse, in June 2009. In 2012, he btained a PhD degree from University of Toulouse. He currently holds a Brunel Postdoctoral Research Fellowship in Statistics, as well as a French Ministère de la Défense Postdoctoral Fellowship, at the School of Mathematics of the University of Bristol. His research activities lie at the intersection of statistical image processing and computational statistics, with a particular interest in Bayesian models and methods for inverse problems. He is particularly interested in Markov Chain Monte Carlo (MCMC) methods for inference in very high dimensional (and often doubly intractable) models, as well as in new connections between MCMC and optimisation methods.School of Mathematics, University of Bristol, UK