Deep Graph Attention Networks

Jun Kato, Airi Mita, Keita Gobara +1

Graphs are useful for representing various realworld objects. However, graph neural networks (GNNs) tend to suffer from over-smoothing, where the representations of nodes of differ…