SAM as an Optimal Relaxation of Bayes

Author name not available (Why is that?)

Publication date: 4 October 2022

Abstract: Sharpness-aware minimization (SAM) and related adversarial deep-learning methods can drastically improve generalization, but their underlying mechanisms are not yet fully understood. Here, we establish SAM as a relaxation of the Bayes objective where the expected negative-loss is replaced by the optimal convex lower bound, obtained by using the so-called Fenchel biconjugate. The connection enables a new Adam-like extension of SAM to automatically obtain reasonable uncertainty estimates, while sometimes also improving its accuracy. By connecting adversarial and Bayesian methods, our work opens a new path to robustness.

Has companion code repository: https://github.com/team-approx-bayes/bayesian-sam

This page was built for publication: SAM as an Optimal Relaxation of Bayes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6412881)