Research Seminar - Leonardo Santoro

The Institute for Statistics and Mathematics is pleased to invite you to the next research seminar, which will take place on campus:

Leonardo Santoro ( Institute of Mathematics, EPFL)
Kernel-Based Generative Sampling via Lifted Gaussian Transport

Wednesday, November 11, 2026, at 4:15 p.m., in Building D4, Room D4.0.039.

(PLEASE NOTE THE DIFFERENT START TIME AND LOCATION!)

Abstract: Kernel methods offer a rich and powerful way to represent data distributions, with their potential extensively demonstrated in discriminative tasks such as testing and classification. In this talk, we explore how Gaussian approximations in a reproducing kernel Hilbert space yield a parametric yet infinite-dimensional representation of data distributions, endowed with a rich geometric structure. Building on this representation, we introduce Lifted Gaussian Transport, a training-free framework for generative sampling that combines kernel methods, Riemannian geometry, and Bures–Wasserstein optimal transport. The resulting procedure is explicit and interpretable, offering a setting to investigate the interplay between geometric approximation, regularization, and generalization.

For further information and the seminar schedule, please see:
https://www.wu.ac.at/en/statmath/research/research-seminar