两篇新研究论文深入探讨了 Sharpness-Aware Minimization (SAM) 的复杂性，SAM 是一种流行的深度学习训练技术。第一篇论文分析了 SAM 在鞍点附近的收敛不稳定性，理论上证明它可以成为一个吸引子，并且动量和批次大小可能对缓解此问题至关重要。第二篇论文为 SAM 专门引入了自适应的 Polyak 型步长调度器，旨在减少对大量学习率调整的需求，同时保持或提高性能。 AI
影响 这些论文为 SAM 提供了理论见解和实际改进，有望实现更稳定、更高效的深度学习模型训练。
排序理由 两篇在 arXiv 上发表的学术论文，讨论了机器学习优化技术的理论方面和改进。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2301.06308v2 Announce Type: replace-cross Abstract: Sharpness-aware minimization (SAM) is a training method that seeks to find flat minima in deep learning, resulting in state-of-the-art performance across various domains. Instead of minimizing the loss of the current weigh…
arXiv:2606.01827v1 Announce Type: cross Abstract: Sharpness-Aware Minimization (SAM) has established itself as a powerful and widely adopted optimizer for training machine learning models. By explicitly minimizing the sharpness of the loss landscape, SAM often improves generaliza…
Sharpness-Aware Minimization (SAM) has established itself as a powerful and widely adopted optimizer for training machine learning models. By explicitly minimizing the sharpness of the loss landscape, SAM often improves generalization while delivering strong empirical performance…