Mitigating Degree Bias Adaptively with Hard-to-Learn Nodes in Graph Contrastive Learning

TL;DR

This paper introduces HAR, a novel contrastive loss that adaptively reweights pairs based on hardness, effectively reducing degree bias in GNNs for node classification, validated by theoretical analysis and experiments.

Contribution

It proposes HAR, a new contrastive loss that adaptively weights pairs, and SHARP, a framework extending HAR to various scenarios, improving degree bias mitigation in GNNs.

Findings

SHARP outperforms baselines on four datasets.

HAR effectively mitigates degree bias.

Theoretical analysis supports HAR's effectiveness.

Abstract

Graph Neural Networks (GNNs) often suffer from degree bias in node classification tasks, where prediction performance varies across nodes with different degrees. Several approaches, which adopt Graph Contrastive Learning (GCL), have been proposed to mitigate this bias. However, the limited number of positive pairs and the equal weighting of all positives and negatives in GCL still lead to low-degree nodes acquiring insufficient and noisy information. This paper proposes the Hardness Adaptive Reweighted (HAR) contrastive loss to mitigate degree bias. It adds more positive pairs by leveraging node labels and adaptively weights positive and negative pairs based on their learning hardness. In addition, we develop an experimental framework named SHARP to extend HAR to a broader range of scenarios. Both our theoretical analysis and experiments validate the effectiveness of SHARP. The experimental results across four datasets show that SHARP achieves better performance against baselines at both global and degree levels.