神经网络在稳定性边缘的对称性破缺分析

作者 PulseAugur 编辑部 · [1 个来源] · 2026-11-05 05:00

研究人员分析了神经网络中的连续对称性如何受到梯度流的影响，特别是在稳定性边缘。他们发现，有限步数的梯度下降会破坏这些守恒定律，导致围绕中心流动力学所描述的更平滑轨迹的振荡。该研究探讨了曲率校正对这些动力学的影响，展示了它如何诱导由临界参数方向上的振荡协方差决定的漂移。这种分析应用于标量两层模型、两层矩阵分解和两层ReLU网络，揭示了在稳定性边缘哪些守恒定律被保留，哪些被破坏。 AI

影响为神经网络训练动力学提供了理论见解，可能为未来的模型架构和优化技术提供信息。

排序理由关于神经网络动力学的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

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

本文如何被排名

Signal score: 26 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 关于神经网络动力学的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, other; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Naima Elosegui Borras, Georgios Arvanitidis · 2026-11-05 05:00

通过中心流分析在稳定边缘实现对称性破缺

arXiv:2611.06718v1 Announce Type: new Abstract: Continuous symmetries in neural networks lead to conserved quantities that are invariant under gradient flow, such as layer balancedness in linear and ReLU networks. Finite-step gradient descent need not preserve these laws, particu…

本文如何被排名

报道来源 [1]

通过中心流分析在稳定边缘实现对称性破缺

相关话题