Heterophily-aware graph attention network

Junfu Wang, Yuanfang Guo, Liang Yang, Yunhong Wang

Abstract: Highlights•We are working for breaking the limitations of graph attention networks to handle heterophilic graphs.•We empirically reveal the benefits of modeling the edge heterophily.•We propose a heterophily-aware attention scheme to adaptively assign weights to different heterophilic types.•We propose a novel Heterophily-Aware Graph Attention Network (HA-GAT).