Negative as Positive: Enhancing Out-of-distribution Generalization for Graph Contrastive Learning

TL;DR

This paper introduces a novel approach in graph contrastive learning that treats semantically similar cross-domain negative pairs as positives, significantly enhancing out-of-distribution generalization.

Contribution

The work proposes the 'Negative as Positive' strategy to improve OOD generalization in GCL by reclassifying certain negative pairs as positives based on semantic similarity.

Findings

Substantial improvement in OOD generalization across multiple datasets.

Traditional InfoNCE optimization limits cross-domain pairs to negatives, hindering OOD performance.

Reclassifying similar negatives as positives enhances domain invariance.

Abstract

Graph contrastive learning (GCL), standing as the dominant paradigm in the realm of graph pre-training, has yielded considerable progress. Nonetheless, its capacity for out-of-distribution (OOD) generalization has been relatively underexplored. In this work, we point out that the traditional optimization of InfoNCE in GCL restricts the cross-domain pairs only to be negative samples, which inevitably enlarges the distribution gap between different domains. This violates the requirement of domain invariance under OOD scenario and consequently impairs the model's OOD generalization performance. To address this issue, we propose a novel strategy "Negative as Positive", where the most semantically similar cross-domain negative pairs are treated as positive during GCL. Our experimental results, spanning a wide array of datasets, confirm that this method substantially improves the OOD generalization performance of GCL.