Expressive GNN

Expressive Graph Neural Networks (GNNs) aim to overcome the limitations of standard GNNs, which often struggle to capture complex graph structures and relationships. Current research focuses on developing GNN architectures that surpass the expressiveness of the Weisfeiler-Lehman test, exploring approaches like higher-order message passing, tensor decomposition for high-order interactions, and novel isomorphism testing methods. These advancements are crucial for improving the accuracy and applicability of GNNs in various domains, including materials science and molecular modeling, where capturing intricate structural details is essential for accurate predictions.

Papers