Chris Nemeth - Multimodal sampling with pseudo-extended MCMC
|Starts:||12:00 8 Oct 2019|
|Ends:||13:00 8 Oct 2019|
|What is it:||Seminar|
|Organiser:||Department of Mathematics|
|Who is it for:||University staff, External researchers, Current University students|
|Speaker:||Dr Chris Nemeth|
Join us for this research seminar, part of the SQUIDS (Statistics, quantification of uncertainty, inverse problems and data science) seminar series.
Abstract: Sampling from posterior distributions using Markov chain Monte Carlo (MCMC) methods can require an exhaustive number of iterations, particularly when the posterior is multi-modal as the MCMC sampler can become trapped in a local mode for a large number of iterations. In this paper, we introduce the pseudo-extended MCMC method as a simple approach for improving the mixing of the MCMC sampler for multi-modal posterior distributions. The pseudo-extended method augments the state-space of the posterior using pseudo-samples as auxiliary variables. On the extended space, the modes of the posterior are connected, which allows the MCMC sampler to easily move between well-separated posterior modes. We demonstrate that the pseudo-extended approach delivers improved MCMC sampling over the Hamiltonian Monte Carlo algorithm on multi-modal posteriors, including Boltzmann machines and models with sparsity-inducing priors. A paper related to this talk can be found here: https://arxiv.org/abs/1708.05239
Speaker Bio: Chris Nemeth (http://www.lancs.ac.uk/~nemeth/) is a lecturer and EPSRC fellow in Statistical Learning in the Department of Mathematics and Statistics at Lancaster University. He completed his PhD in Statistics and Operational in 2014 through the STOR-i Centre for Doctoral Training (https://www.lancaster.ac.uk/stor-i/) and prior to this he was an undergraduate student in Mathematics at the University of Manchester. His core research interests are in computational statistics and machine learning with a focus on Markov chain Monte Carlo algorithms, sequential Monte Carlo and Gaussian processes. Currently, his research is focused on developing scalable Bayesian methods for analysing large-scale data using stochastic gradient Langevin algorithms and related techniques. He is also currently working on the application of data science techniques for tackling environmental science challenges as part of the EPSRC-funded data science for the natural environment project (https://www.lancaster.ac.uk/data-science-of-the-natural-environment/).
Dr Chris Nemeth
Role: Lecturer in Statistical Learning
Organisation: Lancaster University
Travel and Contact Information
Frank Adams 1
Alan Turing Building