On the Impact of Feature Heterophily on Link Prediction with Graph Neural Networks

TL;DR

This paper investigates how feature heterophily affects graph neural network performance in link prediction tasks, providing theoretical insights and empirical evidence to improve model design for diverse network structures.

Contribution

It introduces a formal framework for heterophilic link prediction, analyzes encoder-decoder adaptations, and proposes design improvements for better GNN performance in heterophilic settings.

Findings

Learnable decoders improve link prediction accuracy.

Separating ego- and neighbor-embeddings enhances message passing.

Empirical results confirm the theoretical analysis.

Abstract

Heterophily, or the tendency of connected nodes in networks to have different class labels or dissimilar features, has been identified as challenging for many Graph Neural Network (GNN) models. While the challenges of applying GNNs for node classification when class labels display strong heterophily are well understood, it is unclear how heterophily affects GNN performance in other important graph learning tasks where class labels are not available. In this work, we focus on the link prediction task and systematically analyze the impact of heterophily in node features on GNN performance. Theoretically, we first introduce formal definitions of homophilic and heterophilic link prediction tasks, and present a theoretical framework that highlights the different optimizations needed for the respective tasks. We then analyze how different link prediction encoders and decoders adapt to varying levels of feature homophily and introduce designs for improved performance. Our empirical analysis on a variety of synthetic and real-world datasets confirms our theoretical insights and highlights the importance of adopting learnable decoders and GNN encoders with ego- and neighbor-embedding separation in message passing for link prediction tasks beyond homophily.