Course: STAT 41540
Title: Topics in Advanced Bayesian Methodology
Instructor(s): Aaron Schein
Class Schedule: TR 11:00 AM-12:20 PM in Jones 303
Description: This course will explore topics in advanced Bayesian methodology, particularly around modern variational inference. The course will begin with a review of "classical" variational inference in exponential family models using mean-field approximations. It will then dive into recent advances in generic and scalable VI alternatives such as variational autoencoders, amortized inference, normalizing flows, and diffusion models, among other topics. The course will run like a seminar and feature a mix of lectures and student-run paper presentations. Students will also be responsible for a final project that applies or extends the methodology covered in course to applied problems of their choosing. Prerequisites include STAT 34800 (or an equivalent course on Bayesian statistics and probabilistic graphical models) and some experience in Python is encouraged.