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

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Zixu, Xu, Bingbing, Yuan, Yige, Shen, Huawei, Cheng, Xueqi
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929358385446912
author Wang, Zixu
Xu, Bingbing
Yuan, Yige
Shen, Huawei
Cheng, Xueqi
author_facet Wang, Zixu
Xu, Bingbing
Yuan, Yige
Shen, Huawei
Cheng, Xueqi
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2405_16224
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Negative as Positive: Enhancing Out-of-distribution Generalization for Graph Contrastive Learning
Wang, Zixu
Xu, Bingbing
Yuan, Yige
Shen, Huawei
Cheng, Xueqi
Machine Learning
Artificial Intelligence
I.2
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.
title Negative as Positive: Enhancing Out-of-distribution Generalization for Graph Contrastive Learning
topic Machine Learning
Artificial Intelligence
I.2
url https://arxiv.org/abs/2405.16224