Neural network generalization near interpolation analyzed via statistical mechanics · 2 sources tracked

作者 PulseAugur 编辑部 · [2 个来源] · 2026-07-27 04:00

两篇新的arXiv论文探讨了具有广泛宽度的浅层神经网络在插值阈值附近的泛化能力。该研究使用统计力学分析了这些网络，揭示了在泛化误差独立于权重分布的通用阶段和泛化误差依赖于权重分布的专业化阶段之间存在一个相变。研究结果表明，虽然在插值附近存在高度预测性的解，但由于统计-计算差距，实际算法可能难以找到它们。 AI

影响为神经网络行为提供了理论见解，可能为未来的模型架构和训练策略提供信息。

排序理由两篇在arXiv上发表的学术论文，详细介绍了对神经网络泛化能力的理论分析。

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 →

Neural network generalization near interpolation analyzed via statistical mechanics · 2 sources tracked

报道来源 [2]

arXiv stat.ML TIER_1 English(EN) · Jean Barbier, Francesco Camilli, Minh-Toan Nguyen, Mauro Pastore, Rudy Skerk · 2026-07-27 04:00

Optimal generalisation and learning transition in extensive-width shallow neural networks near interpolation

arXiv:2501.18530v3 Announce Type: replace Abstract: We consider a teacher-student model of supervised learning with a fully-trained two-layer neural network whose width $k$ and input dimension $d$ are large and proportional. We provide an effective theory for approximating the Ba…
arXiv stat.ML TIER_1 English(EN) · Jean Barbier, Francesco Camilli, Minh-Toan Nguyen, Mauro Pastore, Rudy Skerk · 2026-07-27 04:00

Statistical mechanics of extensive-width Bayesian neural networks near interpolation

arXiv:2505.24849v2 Announce Type: replace Abstract: For three decades statistical mechanics has been providing a framework to analyse neural networks. However, the theoretically tractable models, e.g., perceptrons, random features models and kernel machines, or multi-index models…

报道来源 [2]

Optimal generalisation and learning transition in extensive-width shallow neural networks near interpolation

Statistical mechanics of extensive-width Bayesian neural networks near interpolation

相关实体

相关话题