ES-GNN: Generalizing Graph Neural Networks Beyond Homophily with Edge Splitting

arXiv:2205.13700v3 Announce Type: replace
Abstract: While Graph Neural Networks (GNNs) have achieved enormous success in multiple graph analytical tasks, modern variants mostly rely on the strong inductive bias of homophily. However, real-world networks typically exhibit both homophilic and heterophilic linking patterns, wherein adjacent nodes may share dissimilar attributes and distinct labels. Therefore, GNNs smoothing node proximity holistically may aggregate both task-relevant and irrelevant (even harmful) information, limiting their ability to generalize to heterophilic graphs and potentially causing non-robustness. In this …

abstract arxiv beyond bias cs.lg edge gnn gnns graph graph neural networks however inductive modern multiple networks neural networks nodes patterns success tasks type variants world