Graph contrastive learning with node-level accurate difference

Graph contrastive learning (GCL) has attracted extensive research interest due to its powerful ability to capture latent structural and semantic information of graphs in a self-supervised manner. Existing GCL methods commonly adopt predefined graph augmentations to generate two contrastive views. Su...

Full description

Saved in:
Bibliographic Details
Main Authors: Pengfei Jiao, Kaiyan Yu, Qing Bao, Ying Jiang, Xuan Guo, Zhidong Zhao
Format: Article
Language:English
Published: KeAi Communications Co. Ltd. 2025-03-01
Series:Fundamental Research
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2667325824003455
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850064123315879936
author Pengfei Jiao
Kaiyan Yu
Qing Bao
Ying Jiang
Xuan Guo
Zhidong Zhao
author_facet Pengfei Jiao
Kaiyan Yu
Qing Bao
Ying Jiang
Xuan Guo
Zhidong Zhao
author_sort Pengfei Jiao
collection DOAJ
description Graph contrastive learning (GCL) has attracted extensive research interest due to its powerful ability to capture latent structural and semantic information of graphs in a self-supervised manner. Existing GCL methods commonly adopt predefined graph augmentations to generate two contrastive views. Subsequently, they design a contrastive pretext task between these views with the goal of maximizing their agreement. These methods assume the augmented graph can fully preserve the semantics of the original. However, typical data augmentation strategies in GCL, such as random edge dropping, may alter the properties of the original graph. As a result, previous GCL methods overlooked graph differences, potentially leading to difficulty distinguishing between graphs that are structurally similar but semantically different. Therefore, we argue that it is necessary to design a method that can quantify the dissimilarity between the original and augmented graphs to more accurately capture the relationships between samples. In this work, we propose a novel graph contrastive learning framework, named Accurate Difference-based Node-Level Graph Contrastive Learning (DNGCL), which helps the model distinguish similar graphs with slight differences by learning node-level differences between graphs. Specifically, we train the model to distinguish between original and augmented nodes via a node discriminator and employ cosine dissimilarity to accurately measure the difference between each node. Furthermore, we employ multiple types of data augmentation commonly used in current GCL methods on the original graph, aiming to learn the differences between nodes under different augmentation strategies and help the model learn richer local information. We conduct extensive experiments on six benchmark datasets and the results show that our DNGCL outperforms most state-of-the-art baselines, which strongly validates the effectiveness of our model.
format Article
id doaj-art-a098c4449f534fd2a8b7a538d4a67e9f
institution DOAJ
issn 2667-3258
language English
publishDate 2025-03-01
publisher KeAi Communications Co. Ltd.
record_format Article
series Fundamental Research
spelling doaj-art-a098c4449f534fd2a8b7a538d4a67e9f2025-08-20T02:49:23ZengKeAi Communications Co. Ltd.Fundamental Research2667-32582025-03-015281882910.1016/j.fmre.2024.06.013Graph contrastive learning with node-level accurate differencePengfei Jiao0Kaiyan Yu1Qing Bao2Ying Jiang3Xuan Guo4Zhidong Zhao5School of Cyberspace, Hangzhou Dianzi University, Hangzhou 310018, China; Data Security Governance Zhejiang Engineering Research Center, Hangzhou Dianzi University, Hangzhou 310018, ChinaSchool of Cyberspace, Hangzhou Dianzi University, Hangzhou 310018, ChinaCorresponding author.; School of Cyberspace, Hangzhou Dianzi University, Hangzhou 310018, ChinaSchool of Cyberspace, Hangzhou Dianzi University, Hangzhou 310018, ChinaCollege of Intelligence and Computing, Tianjin University, Tianjin 300350, ChinaSchool of Cyberspace, Hangzhou Dianzi University, Hangzhou 310018, ChinaGraph contrastive learning (GCL) has attracted extensive research interest due to its powerful ability to capture latent structural and semantic information of graphs in a self-supervised manner. Existing GCL methods commonly adopt predefined graph augmentations to generate two contrastive views. Subsequently, they design a contrastive pretext task between these views with the goal of maximizing their agreement. These methods assume the augmented graph can fully preserve the semantics of the original. However, typical data augmentation strategies in GCL, such as random edge dropping, may alter the properties of the original graph. As a result, previous GCL methods overlooked graph differences, potentially leading to difficulty distinguishing between graphs that are structurally similar but semantically different. Therefore, we argue that it is necessary to design a method that can quantify the dissimilarity between the original and augmented graphs to more accurately capture the relationships between samples. In this work, we propose a novel graph contrastive learning framework, named Accurate Difference-based Node-Level Graph Contrastive Learning (DNGCL), which helps the model distinguish similar graphs with slight differences by learning node-level differences between graphs. Specifically, we train the model to distinguish between original and augmented nodes via a node discriminator and employ cosine dissimilarity to accurately measure the difference between each node. Furthermore, we employ multiple types of data augmentation commonly used in current GCL methods on the original graph, aiming to learn the differences between nodes under different augmentation strategies and help the model learn richer local information. We conduct extensive experiments on six benchmark datasets and the results show that our DNGCL outperforms most state-of-the-art baselines, which strongly validates the effectiveness of our model.http://www.sciencedirect.com/science/article/pii/S2667325824003455Graph neural networkGraph contrastive learningAccurate difference measureNode representation learningPretext task design
spellingShingle Pengfei Jiao
Kaiyan Yu
Qing Bao
Ying Jiang
Xuan Guo
Zhidong Zhao
Graph contrastive learning with node-level accurate difference
Fundamental Research
Graph neural network
Graph contrastive learning
Accurate difference measure
Node representation learning
Pretext task design
title Graph contrastive learning with node-level accurate difference
title_full Graph contrastive learning with node-level accurate difference
title_fullStr Graph contrastive learning with node-level accurate difference
title_full_unstemmed Graph contrastive learning with node-level accurate difference
title_short Graph contrastive learning with node-level accurate difference
title_sort graph contrastive learning with node level accurate difference
topic Graph neural network
Graph contrastive learning
Accurate difference measure
Node representation learning
Pretext task design
url http://www.sciencedirect.com/science/article/pii/S2667325824003455
work_keys_str_mv AT pengfeijiao graphcontrastivelearningwithnodelevelaccuratedifference
AT kaiyanyu graphcontrastivelearningwithnodelevelaccuratedifference
AT qingbao graphcontrastivelearningwithnodelevelaccuratedifference
AT yingjiang graphcontrastivelearningwithnodelevelaccuratedifference
AT xuanguo graphcontrastivelearningwithnodelevelaccuratedifference
AT zhidongzhao graphcontrastivelearningwithnodelevelaccuratedifference