<ul><li>Recent studies have focused on improving out-of-distribution generalization of Graph Neural Networks (GNNs) in real-world scenarios.</li><li>A new method called PrunE has been proposed to address OOD challenges by eliminating spurious edges in graphs.</li><li>PrunE uses two regularization terms to prune spurious edges and retain the invariant subgraph for better OOD generalization.</li><li>Theoretical analysis and experiments have shown that PrunE outperforms previous methods in achieving superior OOD performance.</li></ul>

Pruning Spurious Subgraphs for Graph Out-of-Distribtuion Generalization

Discover more