Weisfeiler Leman Graph Isomorphism with Molecular Graphs

Graph Neural Networks (GNNs) represent a significant advancement in neural network architectures, specifically tailored to learn vector representations of nodes and entire graphs within a supervised, end-to-end learning framework. However, after careful studies from several authors it has been shown that capabilities of GNNs are in relation to the one-dimensional Weisfeiler-Leman (1-WL) graph isomorphism heuristic. GNNs exhibit an expressive power equivalent to that of the 1-WL heuristic in discerning non-isomorphic subgraphs.