Seminar: David K. Duvenaud - University of Toronto

Date December 14, 2022
Author Hrvoje Stojic

A farewell to GPs: A roadmap for ultra-scalable simulation and inference in stochastic PDEs


This talk sketches a plan for constructing highly-scalable approximate inference schemes in large spatiotemporal models, such as weather or molecular dynamics simulations. Specifically, we'll show how existing heuristics for scaling physical models such as coarse grids or mutli-scale temporal models could be learned automatically by adding auxiliary variables to an approximate variational posterior. We'll also demonstrate a new contribution to parallelizing adaptive SPDE solvers, allowing stateless sampling of entire Brownian sheets of any dimension.



Related articles

Seminar: Arno Solin - Aalto University

Seminar: Vincent Adam - Secondmind & Aalto University

Seminar: Laurence Aitchison - University of Bristol

Seminar: Arthur Gretton - University College London