arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.09308cs.LGmath.GRstat.ML
收藏

论几何图神经网络的表达能力

On the Expressive Power of Geometric Graph Neural Networks

University of Cambridge(剑桥大学);Qualcomm AI Research(高通人工智能研究中心)

Chaitanya K. Joshi, Cristian Bodnar, Simon V. Mathis, Taco Cohen, Pietro Liò

AI总结:

本文提出尊重置换、旋转、反射和平移对称性的几何 WL 测试 GWL,用以刻画几何 GNN 的表达能力,并阐明不变层、等变层、高阶张量与标量化对区分几何图及通用逼近的作用。

摘要:

图神经网络(GNN)的表达能力已通过 Weisfeiler-Leman(WL)图同构测试得到广泛研究。然而,标准 GNN 和 WL 框架并不适用于嵌入在欧几里得空间中的几何图,例如生物分子、材料和其他物理系统。在本工作中,我们提出了一种几何版本的 WL 测试(GWL),用于区分几何图,同时尊重底层的物理对称性:置换、旋转、反射和平移。我们使用 GWL,从区分几何图的角度刻画对物理对称性不变或等变的几何 GNN 的表达能力。GWL 揭示了关键设计选择如何影响几何 GNN 的表达能力:(1)不变层表达能力有限,因为它们无法区分一跳邻域相同的几何图;(2)等变层通过将几何信息传播到局部邻域之外,可区分更大一类图;(3)高阶张量和标量化使几何 GNN 具备最大表达能力;(4)GWL 基于区分的视角等价于通用逼近。补充我们结果的合成实验可在 https://github.com/chaitjo/geometric-gnn-dojo 获取。

英文摘要:

The expressive power of Graph Neural Networks (GNNs) has been studied extensively through the Weisfeiler-Leman (WL) graph isomorphism test. However, standard GNNs and the WL framework are inapplicable for geometric graphs embedded in Euclidean space, such as biomolecules, materials, and other physical systems. In this work, we propose a geometric version of the WL test (GWL) for discriminating geometric graphs while respecting the underlying physical symmetries: permutations, rotation, reflection, and translation. We use GWL to characterise the expressive power of geometric GNNs that are invariant or equivariant to physical symmetries in terms of distinguishing geometric graphs. GWL unpacks how key design choices influence geometric GNN expressivity: (1) Invariant layers have limited expressivity as they cannot distinguish one-hop identical geometric graphs; (2) Equivariant layers distinguish a larger class of graphs by propagating geometric information beyond local neighbourhoods; (3) Higher order tensors and scalarisation enable maximally powerful geometric GNNs; and (4) GWL's discrimination-based perspective is equivalent to universal approximation. Synthetic experiments supplementing our results are available at \url{https://github.com/chaitjo/geometric-gnn-dojo}

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑