Summary of On the Duality Between Sharpness-aware Minimization and Adversarial Training, by Yihao Zhang et al.
On the Duality Between Sharpness-Aware Minimization and Adversarial Training
by Yihao Zhang, Hangzhou He, Jingyu Zhu, Huanran Chen, Yifei Wang, Zeming Wei
First submitted to arxiv on: 23 Feb 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Optimization and Control (math.OC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Medium Difficulty Summary: This paper explores the potential of Sharpness-Aware Minimization (SAM) to enhance adversarial robustness, a property previously associated with Adversarial Training (AT). SAM was designed to improve clean accuracy by perturbing model weights during training. However, this study reveals that using SAM alone can also improve adversarial robustness without sacrificing clean accuracy. The authors provide empirical and theoretical insights into how SAM learns more robust features, demonstrating its ability to enhance adversarial robustness. This work sheds light on the potential of SAM as a substitute for AT when prioritizing accuracy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty Summary: Researchers have found a new way to make machines more resistant to attacks that try to trick them. Instead of using a method called Adversarial Training (AT), they used another approach called Sharpness-Aware Minimization (SAM). Surprisingly, SAM worked just as well as AT at keeping the machine safe from these attacks. The study explains how SAM helps the machine learn more robust features and shows that it can be used instead of AT when accuracy is more important. |
Keywords
* Artificial intelligence * Sam
