PulseAugur
实时 16:11:45
English(EN) Broken Ergodicity and the Violation of the Fluctuation-Dissipation Theorem Lead to Generalization Beyond Overfitting in Machine Learning

Machine learning double descent linked to superconducting transition physics

一篇新论文探讨了机器学习中的“双下降”现象,即神经网络的泛化能力在其复杂度超过训练数据量时仍能继续提高。研究人员利用动力学平均场理论证明,这种行为源于控制训练过程的随机场论中的一个相变。该相变以遍历性破坏导致涨落耗散定理失效为特征,网络的泛化能力与超导转变的伦敦模型相似。 AI

影响 这项研究为理解复杂神经网络中的泛化提供了理论框架,可能指导未来的模型开发。

排序理由 Academic paper detailing a theoretical finding in machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

在 Hugging Face Daily Papers 阅读 →

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

Machine learning double descent linked to superconducting transition physics

报道来源 [1]

  1. Hugging Face Daily Papers TIER_1 English(EN) ·

    不可靠的遍历性与涨落-耗散定理的违反导致机器学习中的过拟合泛化

    The remarkable ability of modern neural networks to generalize improves with increasing network capacity, even when the number of model parameters or effective degrees of freedom exceeds the number of training data points. This phenomenon is all the more surprising given that gen…