On the Expressive Power of Geometric Graph Neural Networks

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

Keywords: geometric deep learning, graph neural networks, graph isomorphism

TL;DR: We propose a Geometric Weisfeiler-Leman test to study the expressive power of geometric graph neural networks.

Abstract: We propose a geometric version of the Weisfeiler-Leman graph isomorphism 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 Graph Neural Networks (GNNs) that are invariant or equivariant to physical symmetries in terms of the classes of geometric graphs they can distinguish. This allows us to formalise the advantages of equivariant GNNs over invariant GNNs: equivariant layers have greater expressive power as they enable propagating geometric information beyond local neighbourhoods, while invariant layers only reason locally via scalars and cannot discriminate geometric graphs with different non-local properties.

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2301.09308/code)