arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

神经网络的分形小工具:窄宽度情形的复杂性

Fractal Gadgets for Neural Networks: The Complexity of the Narrow Regime

Olivier Bournez, Johanne Cohen, Laura Cohen, Adrian Wurm

arXiv 2610.06092首次发表:更新:

发表机构

Institut Polytechnique de Paris, Ecole Polytechnique; LISN, CNRS, Université Paris-Saclay; CentraleSupelec, Université Paris-Saclay; Computer Science Institute, BTU Cottbus-Senftenberg(巴黎理工学院,巴黎综合理工学院; LISN,法国国家科学研究中心,巴黎萨克雷大学; 中央理工-高等电力学院,巴黎萨克雷大学; 科特布斯-森夫滕贝格技术大学计算机研究所)

机构由 AI 辅助整理,请以论文原文为准。

AI 中文总结

研究窄宽度ReLU网络验证的复杂性,通过分形预处理小工具证明宽度4时NP完全,宽度3在离散输入下NP完全,并给出计数与近似等进一步结果。

AI 中文摘要

我们研究深度窄ReLU神经网络的验证问题:给定一个宽度有界、在[0,1]上计算分段仿射映射的网络,是否存在某个输入满足规定的输出约束?经典的ReLU验证NP困难性证明使用每个布尔变量一个神经元,且对小的常数宽度网络无话可说,而宽度1的网络易于验证。我们证明,对于[0,1]中的任意输入,宽度为4的ReLU网络验证是NP完全的。当输入被限制在一个自然的离散编码集时,NP完全性在宽度3时已经成立。结合宽度1时的多项式时间可判定性,这仅在编码集上留下宽度2,以及在[0,1]上留下宽度2和3的问题未解决。技术核心是一个分形预处理小工具:一个宽度为2的ReLU子网络,其迭代恰好在一个包含2^n个点的有限类Cantor子集附近消失。它将[0,1]上连续函数的验证归结为对2^n个离散点的验证,而不增加宽度,并且是宽度有界困难性归约所缺失的要素。同样的构造在编码集上的宽度3情形产生了进一步的结果:通用问题是coNP完全的,计数零点问题是#P完全的,多数变体是PP完全的,以及在由Max-3Sat继承的常数间隙内近似最小输出是NP困难的。NP、coNP和不可近似性结果在宽度4时提升到整个[0,1];提升计数和多数问题,以及在宽度3时的提升,仍然是开放的。

英文摘要

We study the verification problem for deep narrow ReLU neural networks: given a network of bounded width computing a piecewise-affine map on [0,1], does some input satisfy a prescribed output constraint? Classical NP-hardness proofs for ReLU verification use one neuron per Boolean variable and say nothing about networks of small constant width, while width-1 networks are easy to verify. We show that verification of ReLU networks is NP-complete at width 4 for arbitrary inputs in [0,1]. When inputs are restricted to a natural discrete encoding set, NP-completeness already holds at width 3. Together with polynomial-time decidability at width 1, this leaves open only width 2 on the encoding set, and widths 2 and 3 on [0,1]. The technical core is a fractal preprocessing gadget: a width-2 ReLU subnetwork whose iterate vanishes precisely near a finite Cantor-like subset of [0,1] with 2^n points. It reduces verification of a continuous function on [0,1] to verification on 2^n discrete points without increasing the width, and is the missing ingredient for width-bounded hardness reductions. The same construction yields further results at width 3 on the encoding set: the universal problem is coNP-complete, counting zeros is #P-complete, a majority variant is PP-complete, and approximating the minimum output within a constant gap inherited from Max-3Sat is NP-hard. The NP, coNP and inapproximability results lift to all of [0,1] at width 4; lifting counting and majority, and lifting at width 3, remain open.

Comments30 pages, 8 figures, Accepted at FSTTCS 2026

论文原文

arXiv 摘要页 · PDF 原文 · HTML 原文

↑