New LE-SAM method boosts model generalization over traditional SAM

Researchers have introduced Loss-Equated SAM (LE-SAM), a novel approach to enhance generalization in machine learning models. This method addresses a mismatch in Sharpness-Aware Minimization (SAM) by focusing on a fixed loss-space budget rather than a fixed perturbation radius. LE-SAM effectively prioritizes curvature-dominated optimization terms over gradient-norm signals. Experiments show LE-SAM consistently outperforms SAM and its variants, achieving state-of-the-art results on various benchmarks. AI

影响 Introduces a new optimization technique that improves model generalization, potentially leading to more robust AI systems.

排序理由 Academic paper introducing a new method for machine learning optimization. [lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

• SAM

• 论文
• 其他

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →