Probabilistic Inference and Learning with Stein’s Method

Abstract: Stein’s method is a powerful tool from probability theory for bounding the distance between probability distributions. In this talk, I’ll describe how this tool designed to prove central limit theorems can be adapted to assess and improve the quality of practical inference procedures. I’ll highlight applications to Markov chain Monte Carlo sampler selection, goodness-of-fit testing, variational inference, and nonconvex optimization and close with several opportunities for future work.

Lester Mackey (https://web.stanford.edu/~lmackey/) received his PhD from UC Berkeley under the supervision of Michael Jordan. Between 2013 and 2016 he held an Assistant Professorship at Stanford University and is now a Principal Researcher at Microsoft Research and an adjunct professor at Stanford. His work on measuring MCMC sample quality with Stein’s method from 2015 is considered foundational for the field of Stein’s method in ML and opened the door to countless other publications in this area. His own contribution in the field has been immense - he has published articles covering various applications of Stein’s method in ML, including to problems related to computational statistics and statistical testing.