PulseAugur
中
实时 08:22:05
English(EN) Why Shallow Neural Networks Learn Sinusoidal Components First: An Orthogonal Polynomial Dynamics Framework

新框架解释了神经网络为何首先学习低频分量

一个名为正交多项式动力学框架(Orthogonal Polynomial Dynamics Framework)的新理论框架已被开发出来,用于解释浅层神经网络为何优先学习低频分量,这种现象被称为谱偏差(spectral bias)。该框架分析了动力学矩阵 H,它控制着正交多项式基下傅里叶级数系数的梯度下降。该研究旨在提供一个适用于各种激活函数和网络架构的通用解释,以克服先前在理解谱偏差方面的局限性。 AI

影响 为理解和潜在控制神经网络的学习过程提供了理论基础。

排序理由 该条目是一篇学术论文,详细介绍了一个理解神经网络行为的新理论框架。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

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

新框架解释了神经网络为何首先学习低频分量

本文如何被排名

Signal score
17 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该条目是一篇学术论文,详细介绍了一个理解神经网络行为的新理论框架。[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]

  1. arXiv cs.AI TIER_1 English(EN) · Yanyang Xiao ·

    浅层神经网络为何首先学习正弦分量:一个正交多项式动力学框架

    arXiv:2610.21964v1 Announce Type: cross Abstract: Neural networks often learn low-frequency components before high-frequency ones --- a phenomenon called spectral bias or termed F-principle. Many numerical studies have demonstrated this phenomenon, and theoretical studies have so…