How powerful are graph neural networks?

Beyond Weisfeiler-Lehman: using substructures for provably expressive graph neural networks

In this post, I discuss how to design local and computationally efficient provably powerful graph neural networks that are not based on the Weisfeiler-Lehman tests hierarchy.

Michael Bronstein
8 min readJul 3, 2020

--

--

--

DeepMind Professor of AI @Oxford. Serial startupper. ML for graphs, biochemistry, drug design, and animal communication.