AM Seminar: Sliced Optimal Transport: From Acceleration to Bayesian Modeling

Presenter: Enter the speaker’s name, title, and institution at the top of the event

Description: Optimal transport (OT) provides a principled way to match, compare, and interpolate distributions, but its computational cost limits its use at realistic scales. I will present our recent work on reducing that cost and on using OT for statistical modeling and inference. In the first part of the talk, I will briefly introduce optimal transport and sliced optimal transport (SOT). I will then present sliced-regularized OT, which uses a cheap sliced plan as an informative prior for approximating the full plan, yielding an efficient approximate OT solver. In the second part, I will show how SOT enables new applications in statistical inference. In particular, I will introduce distributional determinantal point processes, which use SOT geometry to place repulsive priors on distributions, giving a Bayesian model for clustering distributional data. Applications include image processing and the clustering of single-cell and epilepsy data.

About the speaker: Dr. Khai Nguyen is an Assistant Professor in the Department of Statistics and Data Science at Texas A&M University. He received his Ph.D. in Statistics from the University of Texas at Austin and his B.Sc. in Computer Science from Hanoi University of Science and Technology. His research develops computational optimal transport, particularly sliced methods that make large-scale transport tractable. He brings these tools to problems in machine learning and geometric data processing, as well as to Bayesian approximate inference and nonparametric models for distributional data.

This seminar is hosted by Professor Yunyi Shen.