Graph Contrastive Learning under Heterophily via Graph Filters

TL;DR

This paper introduces HLCL, a graph contrastive learning method that effectively handles heterophilic graphs by using graph filters to differentiate and aggregate node representations, outperforming existing methods.

Contribution

HLCL is the first contrastive learning approach that explicitly distinguishes homophilic and heterophilic subgraphs using graph filters for better node representation learning.

Findings

HLCL outperforms state-of-the-art methods on heterophilic benchmark datasets.

HLCL achieves up to 10% better accuracy than supervised methods on heterophilic graphs.

HLCL effectively leverages low-pass and high-pass filters for improved node embeddings.

Abstract

Graph contrastive learning (CL) methods learn node representations in a self-supervised manner by maximizing the similarity between the augmented node representations obtained via a GNN-based encoder. However, CL methods perform poorly on graphs with heterophily, where connected nodes tend to belong to different classes. In this work, we address this problem by proposing an effective graph CL method, namely HLCL, for learning graph representations under heterophily. HLCL first identifies a homophilic and a heterophilic subgraph based on the cosine similarity of node features. It then uses a low-pass and a high-pass graph filter to aggregate representations of nodes connected in the homophilic subgraph and differentiate representations of nodes in the heterophilic subgraph. The final node representations are learned by contrasting both the augmented high-pass filtered views and the augmented low-pass filtered node views. Our extensive experiments show that HLCL outperforms state-of-the-art graph CL methods on benchmark datasets with heterophily, as well as large-scale real-world graphs, by up to 7%, and outperforms graph supervised learning methods on datasets with heterophily by up to 10%.