English(EN) Functional Degeneracy in Neural Networks: Measurement and Pruning

测量神经网络退化以指导模型压缩

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-01 04:00

研究人员开发了一种量化神经网络功能退化的方法，该方法衡量模型在不影响其性能的情况下可以压缩多少。这是通过计算“行为恢复秩”来实现的，该秩代表恢复训练模型能力所需的行为-Hessian特征方向的数量。研究发现，结构和幅度剪枝方法保留了更多的自由度，这表明功能冗余分布在参数方向上，而不是局限于单个权重或神经元。 AI

影响提供了一个新的指标来理解和潜在地优化神经网络压缩和效率。

排序理由该集群包含一篇详细介绍分析神经网络新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

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

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; 该集群包含一篇详细介绍分析神经网络新方法的学术论文。[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, model release; 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]

arXiv cs.LG TIER_1 English(EN) · Maria Matveev, Pascal Esser, Ayush Bharadwaj, Lucius Bushnaq, Gitta Kutyniok · 2026-09-01 04:00

神经网络中的函数退化：测量与剪枝

arXiv:2608.30741v1 Announce Type: new Abstract: A central question in modern machine learning is how much a trained model can be compressed without changing its behavior, to reduce the memory, compute and energy required to deploy it. To study this, we quantify functional degener…

本文如何被排名

报道来源 [1]

神经网络中的函数退化：测量与剪枝

相关实体

相关话题