PulseAugur
中
实时 21:16:04
English(EN) Deviations from the NTK Regime in CNNs under Joint Depth-Width Scaling

CNNs:深度-宽度缩放影响神经切线核行为

本文研究了在深度和宽度同时缩放时,卷积神经网络(CNNs)中神经切线核(NTK)的行为。研究人员发现,在特定的初始化机制下,NTK色散(衡量核变异性的指标)随着深度与宽度之比(L/C)的增加而增加。此外,即使初始色散较低,在训练过程中发生的核旋转也普遍随着L/C的增加而增加。这些发现表明,L/C比率在CNNs偏离理论无限宽度极限方面起着重要作用,并且初始化和训练稳定性具有不同的缩放行为。 AI

影响 了解深度-宽度缩放如何影响CNNs中的NTK行为,可以为未来改进训练稳定性和性能的架构设计提供信息。

排序理由 该条目是一篇研究论文,详细介绍了对神经网络行为的数值研究。[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 →

CNNs:深度-宽度缩放影响神经切线核行为

本文如何被排名

Signal score
1 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该条目是一篇研究论文,详细介绍了对神经网络行为的数值研究。[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
1 days old
Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

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

    CNNs在联合深度宽度缩放下的NTK机制偏差

    The concentration and training-time constancy of the neural tangent kernel (NTK) depend on the interplay between network width, depth, and initialization. While joint depth-width scaling has been studied in fully-connected, residual, and recurrent networks, its effects in convolu…