
The author proposes a Directional Graph Attention Network (DGAT) to address the limitations of existing Graph Attention Networks (GAT) on heterophilic graphs. By introducing a new class of Laplacian matrices and topology-guided mechanisms, DGAT outperforms state-of-the-art models in node classification tasks.


coremsg

Enhancing-Graph-Attention-Networks-with-Directional-Neighborhood-Attention


Enhancing Graph Attention Networks with Directional Neighborhood Attention


title_rewrite


DGAT enhances GAT performance by incorporating global directional information and topology-guided strategies.


heterophilic-graph-with-directional-neighborhood-attention-for-representation-learning


Heterophilic Graph with Directional Neighborhood Attention for Representation Learning
