Exploring the Effect of Multi-step Ascent in Sharpness-Aware Minimization
Abstract
Recently, Sharpness-Aware Minimization (SAM) has shown state-of-the-art performance by seeking flat minima. To minimize the maximum loss within a neighborhood in the parameter space, SAM uses an ascent step, which perturbs the weights along the direction of gradient ascent with a given radius. While single-step or multi-step can be taken during ascent steps, previous studies have shown that multi-step ascent SAM rarely improves generalization performance. However, this phenomenon is particularly interesting because the multi-step ascent is expected to provide a better approximation of the maximum neighborhood loss. Therefore, in this paper, we analyze the effect of the number of ascent steps and investigate the difference between both single-step ascent SAM and multi-step ascent SAM. We identify the effect of the number of ascent on SAM optimization and reveal that single-step ascent SAM and multi-step ascent SAM exhibit distinct loss landscapes. Based on these observations, we finally suggest a simple modification that can mitigate the inefficiency of multi-step ascent SAM.
Keywords
Cite
@article{arxiv.2302.10181,
title = {Exploring the Effect of Multi-step Ascent in Sharpness-Aware Minimization},
author = {Hoki Kim and Jinseong Park and Yujin Choi and Woojin Lee and Jaewook Lee},
journal= {arXiv preprint arXiv:2302.10181},
year = {2023}
}