English(EN) Parameterized Hardness of Zonotope Containment and Neural Network Verification

新arXiv论文分析神经网络验证的复杂度

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

一篇新发表在arXiv上的论文探讨了神经网络验证的参数化计算复杂度。研究人员证明，对于具有$\ell$层和ReLU激活函数的神经网络，当以输入维度$d$为参数时，确定其正性和满射性等性质是W[$\ell-1$]-难的。这一发现解决了悬而未决的问题，并对计算几何、控制理论和机器人学中的问题（如区域图非包含性）产生了影响。研究还表明，对于常数$d$，近似这些网络中的Lipschitz常数和其他性质是NP-难和W[$\ell-1$]-难的；而对于常数$\\ell$，则是NP-难和W[$t$]-难的。 AI

影响确立了验证ReLU神经网络性质的可处理性的理论极限，影响了未来的算法设计。

排序理由该集群包含一篇详细介绍神经网络验证的理论计算复杂度结果的研究论文。[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: 15 / 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神经网络验证的理论计算复杂度结果的研究论文。[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, 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]

arXiv cs.LG TIER_1 English(EN) · Vincent Froese, Moritz Grillo, Christoph Hertrich, Moritz Stargalla · 2026-09-04 04:00

参数化区域图包含和神经网络验证的硬度

arXiv:2509.22849v3 Announce Type: replace-cross Abstract: Neural networks with ReLU activations are a widely used model in machine learning. It is thus important to have a profound understanding of the properties of the functions computed by such networks. Recently, there has bee…

本文如何被排名

报道来源 [1]

参数化区域图包含和神经网络验证的硬度

相关实体

相关话题