Gradient Estimation with Stochastic Softmax Tricks

Max Paulus, Dami Choi, Daniel Tarlow, Andreas Krause, Chris J. Maddison

Oral presentation: Orals & Spotlights Track 19: Probabilistic/Causality
on 2020-12-09T06:30:00-08:00 - 2020-12-09T06:45:00-08:00
Poster Session 4 (more posters)
on 2020-12-09T09:00:00-08:00 - 2020-12-09T11:00:00-08:00
GatherTown: Probabilistic methods and model efficiency ( Town E4 - Spot C1 )
Join GatherTown
Only iff poster is crowded, join Zoom . Authors have to start the Zoom call from their Profile page / Presentation History.
Abstract: The Gumbel-Max trick is the basis of many relaxed gradient estimators. These estimators are easy to implement and low variance, but the goal of scaling them comprehensively to large combinatorial distributions is still outstanding. Working within the perturbation model framework, we introduce stochastic softmax tricks, which generalize the Gumbel-Softmax trick to combinatorial spaces. Our framework is a unified perspective on existing relaxed estimators for perturbation models, and it contains many novel relaxations. We design structured relaxations for subset selection, spanning trees, arborescences, and others. When compared to less structured baselines, we find that stochastic softmax tricks can be used to train latent variable models that perform better and discover more latent structure.

Preview Video and Chat

Chat is not available.