Neighbor Overlay-Induced Graph Attention Network

TL;DR

This paper introduces NO-GAT, a novel graph attention network that incorporates explicit structural information from overlaid neighbors to improve node representation learning.

Contribution

The paper proposes a neighbor overlay technique for GATs that enhances attention mechanisms with structural cues, outperforming existing models.

Findings

NO-GAT outperforms state-of-the-art models on benchmark datasets.

Incorporating neighbor overlays improves attention coefficient accuracy.

Structural information from overlaid neighbors enhances node representations.

Abstract

Graph neural networks (GNNs) have garnered significant attention due to their ability to represent graph data. Among various GNN variants, graph attention network (GAT) stands out since it is able to dynamically learn the importance of different nodes. However, present GATs heavily rely on the smoothed node features to obtain the attention coefficients rather than graph structural information, which fails to provide crucial contextual cues for node representations. To address this issue, this study proposes a neighbor overlay-induced graph attention network (NO-GAT) with the following two-fold ideas: a) learning favorable structural information, i.e., overlaid neighbors, outside the node feature propagation process from an adjacency matrix; b) injecting the information of overlaid neighbors into the node feature propagation process to compute the attention coefficient jointly. Empirical studies on graph benchmark datasets indicate that the proposed NO-GAT consistently outperforms state-of-the-art models.