\dm_csml_event_details UCL ELLIS

Sharpness-Aware Minimization (SAM) Current Method and Future Directions


Speaker

Hossein Mobahi

Affiliation

Google

Date

Friday, 25 February 2022

Time

17:00-18:00

Location

Zoom

Link

https://ucl.zoom.us/j/96768812815

Event series

Jump Trading/ELLIS CSML Seminar Series

Abstract

In today’s heavily overparameterized models, the value of the training loss provides few guarantees on model generalization ability. Indeed, optimizing only the training loss value, as is commonly done, can easily lead to suboptimal model quality. Motivated by prior work connecting the geometry of the loss landscape and generalization, we introduce a new and effective procedure for instead simultaneously minimizing loss value and loss sharpness. In particular, our procedure, Sharpness-Aware Minimization (SAM), seeks parameters that lie in neighborhoods having uniformly low loss; this formulation results in a min-max optimization problem on which gradient descent can be performed efficiently. We present empirical results showing that SAM improves model generalization across a variety of benchmark datasets (e.g., CIFAR-10, CIFAR-100, ImageNet, finetuning tasks) and models, yielding novel state-of-the-art performance for several. Additionally, we find that SAM natively provides robustness to label noise on par with that provided by state-of-the-art procedures that specifically target learning with noisy labels. Finally, we will discuss possible directions for further research around SAM.

Biography