VASSO: Variance Suppression for Sharpness-Aware Minimization

Bingcong Li, Yilang Zhang, Georgios B. Giannakis · 2025

Read VASSO: Variance Suppression for Sharpness-Aware Minimization by Bingcong Li, Yilang Zhang, Georgios B. Giannakis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-02433.txt.

Description

Bingcong Li, Yilang Zhang, and Georgios B. Giannakis wrote a 2025 machine learning paper improving sharpness-aware minimisation, a training method that seeks flat valleys in the loss surface because flat minima tend to generalise better. The method perturbs weights toward the worst nearby loss before updating.

In practice those perturbations can be too friendly, staying tame and limiting the gain. Their VASSO approach suppresses the variance of the adversary, the perturbation process, so it stays stable and probing. Theory supports the stabilisation claim. Added to standard and efficient variants of the base method, VASSO improved generalisation across vision and language tasks. On top of a cheap variant, it offers a good balance between accuracy gains and computing cost.

Catalog record

CREATORBingcong Li, Yilang Zhang, Georgios B. Giannakis
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE77 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-02433

Browse the full library · About the Tower of Babel