Bamdad Hosseini - Transport maps for conditional simulation in high dimensions
|Starts:||15:00 20 Jul 2022|
|Ends:||16:00 20 Jul 2022|
|What is it:||Seminar|
|Organiser:||Department of Mathematics|
|Who is it for:||University staff, External researchers, Current University students|
|Speaker:||Dr Bamdad Hosseini|
Join us for this research seminar, part of the SQUIDS (Statistics, quantification of uncertainty, inverse problems and data science) seminar series.
Abstract: MCMC algorithms are the gold standard for sampling of high-dimensional probability measures and in particular for Bayesian inference. Their convergence properties and accuracy have been the subject of intense research over the past few decades. However, MCMC methods suffer from certain drawbacks in high-dimensional applications specially as it pertains to inverse problems. In this talk I will discuss an alternative approach to conditional simulation based on transport maps. Broadly speaking, these methods fall under the category of variational inference but are inspired by recent advances in machine learning and uncertainty quantification. I will present some foundational theory as well as numerical experiments that demonstrate the feasibility of these new ideas.
Dr Bamdad Hosseini
Organisation: University of Washington
Travel and Contact Information
Frank Adams 1
Alan Turing Building