Beyond Homophily: Neighborhood Distribution-guided Graph Convolutional Networks

Siqi Liu, Dongxiao He, Zhizhi Yu, Di Jin, Zhiyong Feng

Abstract: Highlights•Similar neighborhood distributions found in homophilic and heterophilic graphs.•New graph convolutional network integrates lower and higher-order neighborhoods.•A novel feature contrastive loss for better node representations.•Experimental results on seven datasets validate the effectiveness of ND-GCN.