两篇新的arXiv论文探讨了深度神经网络在训练过程中的谱动力学。一篇论文引入了“神经低秩滤波”（Neural LoFi）作为理论框架，将分层特征学习理解为一种迭代谱过程。另一篇论文使用动力学平均场理论分析隐藏权重谱的演变方式，预测宽网络中的离群行为和超参数迁移。 AI
影响 这些理论框架为理解深度神经网络的学习方式提供了新的视角，可能指导未来的模型开发和分析。
排序理由 两篇在arXiv上发表的学术论文，提出了理解深度学习的理论框架。
AI 生成摘要 · Google Gemini · 来自 5 个来源。 我们如何撰写摘要 →
Understanding how deep neural networks learn useful internal representations from data remains a central open problem in the theory of deep learning. We introduce Neural Low-Degree Filtering (Neural LoFi), a stylized limit of gradient-based training in which hierarchical feature …
Understanding how deep neural networks learn useful internal representations from data remains a central open problem in the theory of deep learning. We introduce Neural Low-Degree Filtering (Neural LoFi), a stylized limit of gradient-based training in which hierarchical feature …
We study the evolution of hidden-weight spectra in wide neural networks trained by (stochastic) gradient descent. We develop a two-level dynamical mean-field theory (DMFT) that jointly tracks bulk and outlier spectral dynamics for spiked ensembles whose spike directions remain st…
arXiv:2605.13612v1 Announce Type: cross Abstract: Understanding how deep neural networks learn useful internal representations from data remains a central open problem in the theory of deep learning. We introduce Neural Low-Degree Filtering (Neural LoFi), a stylized limit of grad…
arXiv:2605.07870v1 Announce Type: cross Abstract: We study the evolution of hidden-weight spectra in wide neural networks trained by (stochastic) gradient descent. We develop a two-level dynamical mean-field theory (DMFT) that jointly tracks bulk and outlier spectral dynamics for…