PulseAugur
实时 08:50:41
English(EN) A law of robustness for two-layer neural networks with arbitrary weights

研究人员证明了两层神经网络的鲁棒性定律

研究人员证明了具有任意权重的两层神经网络的“鲁棒性定律”,解决了 BubeckLiNagaraj 的一个猜想。该证明适用于 ReLU 等连续分段线性激活函数,它表明在某个阈值以下拟合噪声数据需要特定的 Lipschitz 常数。这一发现很重要,因为它不需要对网络权重的规模进行限制,而这是先前相关证明中存在的局限性。 AI

影响 这项理论工作增进了对神经网络特性的理解,可能为未来的模型设计和分析提供信息。

排序理由 该集群包含一篇详细介绍机器学习理论发现的学术论文。

在 arXiv stat.ML 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

研究人员证明了两层神经网络的鲁棒性定律

报道来源 [2]

  1. arXiv stat.ML TIER_1 English(EN) · Yitzchak Shmalo ·

    A law of robustness for two-layer neural networks with arbitrary weights

    arXiv:2607.07778v1 Announce Type: cross Abstract: Bubeck, Li and Nagaraj conjectured that, for generic data, any two-layer neural network with $m$ neurons that fits $n$ noisy labels must have Lipschitz constant at least of order $\sqrt{n/m}$, with no restriction on the size of th…

  2. arXiv stat.ML TIER_1 English(EN) · Yitzchak Shmalo ·

    A law of robustness for two-layer neural networks with arbitrary weights

    Bubeck, Li and Nagaraj conjectured that, for generic data, any two-layer neural network with $m$ neurons that fits $n$ noisy labels must have Lipschitz constant at least of order $\sqrt{n/m}$, with no restriction on the size of the weights. Bubeck and Sellke proved a universal ve…