English(EN) Scaling Laws from Sequential Feature Recovery: A Solvable Hierarchical Model

新模型解释了从顺序特征学习中产生的标度律

作者 PulseAugur 编辑部 · [2 个来源] · 2026-05-14 08:37

研究人员开发了一个分层模型，解释了多层神经网络中标度律如何从特征学习中产生。该模型利用一种逐层谱算法，该算法顺序地恢复潜在的组合特征，更强的特征可以更早地被检测到，并且需要更少的数据。该方法基于随机矩阵理论，与浅层方法相比，显示出改进的标度性能，并为理解源自尖锐特征学习过渡的平滑标度律提供了理论框架。 AI

影响为理解模型性能如何随数据和复杂性进行标度提供了理论框架。

排序理由该集群包含一篇 arXiv 预印本，详细介绍了理解神经网络中标度律的新理论模型。

在 Hugging Face Daily Papers 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

Hugging Face Daily Papers TIER_1 English(EN) · 2026-05-14 08:37

Scaling Laws from Sequential Feature Recovery: A Solvable Hierarchical Model

We propose a simple mechanism by which scaling laws emerge from feature learning in multi-layer networks. We study a high-dimensional hierarchical target that is a globally high-degree function, but that can be represented by a combination of latent compositional features whose w…
arXiv stat.ML TIER_1 English(EN) · Bruno Loureiro · 2026-05-14 08:37

Scaling Laws from Sequential Feature Recovery: A Solvable Hierarchical Model

We propose a simple mechanism by which scaling laws emerge from feature learning in multi-layer networks. We study a high-dimensional hierarchical target that is a globally high-degree function, but that can be represented by a combination of latent compositional features whose w…

报道来源 [2]

Scaling Laws from Sequential Feature Recovery: A Solvable Hierarchical Model

Scaling Laws from Sequential Feature Recovery: A Solvable Hierarchical Model

相关实体

相关话题