Directed Homophily-Aware Graph Neural Network

TL;DR

DHGNN is a novel graph neural network framework that effectively handles heterophilic and directed graphs by incorporating homophily-aware and direction-sensitive modules, leading to superior performance in node classification and link prediction.

Contribution

The paper introduces DHGNN, a new GNN model that explicitly accounts for directionality and homophily, addressing limitations of existing GNNs on complex directed graphs.

Findings

DHGNN outperforms state-of-the-art methods in node classification and link prediction.

DHGNN improves link prediction accuracy by up to 15.07%.

The gating mechanism captures directional homophily gaps across layers.

Abstract

Graph Neural Networks (GNNs) have achieved significant success in various learning tasks on graph-structured data. Nevertheless, most GNNs struggle to generalize to heterophilic neighborhoods. Additionally, many GNNs ignore the directional nature of real-world graphs, resulting in suboptimal performance on directed graphs with asymmetric structures. In this work, we propose Directed Homophily-aware Graph Neural Network (DHGNN), a novel framework that addresses these limitations by incorporating homophily-aware and direction-sensitive components. DHGNN employs a resettable gating mechanism to adaptively modulate message contributions based on homophily levels and informativeness, and a structure-aware noise-tolerant fusion module to effectively integrate node representations from the original and reverse directions. Extensive experiments on both homophilic and heterophilic directed graph datasets demonstrate that DHGNN outperforms state-of-the-art methods in node classification and link prediction. In particular, DHGNN improves over the best baseline by up to 15.07\% in link prediction. Our analysis further shows that the gating mechanism captures directional homophily gaps and fluctuating homophily across layers, providing deeper insights into message-passing behavior on complex graph structures.