Tyler Maunu (Brandeis): Low-dimensional structure for optimal transport and gradient flows
Please note this event occurred in the past.
September 10, 2026 4:00 pm - 5:00 pm ET
LGRT 1685
Abstract
Abstract:
I will discuss recent work that exploits low-dimensional structure to improve the efficiency and interpretability of methods for optimal transport and sampling. In the first part, I will present Schatten OT, a framework that uses Schatten regularization to promote low-rank structure in transport plans, displacements, and covariances, with both theoretical and empirical support. In the second part, I will describe subspace-projected variants of adaptive Langevin algorithms, which enable faster sampling with low memory overhead. I will finish with a discussion of the extension of these ideas to Laplacian adjusted transport flows.