计算机科学

计算机科学 ›› 2026, Vol. 53 ›› Issue (2): 145-151.doi: 10.11896/jsjkx.250100155

• 数据库&大数据&数据科学 • 上一篇    下一篇

融合对比学习的掩码图自编码器

王新喻, 宋小民, 郑慧明, 彭德中, 陈杰   

  1. 四川大学计算机学院 成都 610065
  • 收稿日期:2025-01-24 修回日期:2025-04-27 发布日期:2026-02-10
  • 通讯作者: 陈杰(chenjie2010@scu.edu.cn)
  • 作者简介:(wangxinyu11@stu.scu.edu.cn)
  • 基金资助:
    国家自然科学基金(62372315);四川省科技计划(2024NSFTD0049,2024ZDZX0004)

Contrastive Learning-based Masked Graph Autoencoder

WANG Xinyu, SONG Xiaomin, ZHENG Huiming, PENG Dezhong, CHEN Jie   

  1. College of Computer Science,Sichuan University,Chengdu 610065,China
  • Received:2025-01-24 Revised:2025-04-27 Online:2026-02-10
  • About author:WANG Xinyu,born in 2001,postgra-duate.His main research interest is graph neural network.
    CHEN Jie,born in 1982,Ph.D,associate professor.His main research interests include machine learning,big data ana-lysis and artificial intelligence.
  • Supported by:
    National Natural Science Foundation of China(62372315) and Sichuan Provincial Science and Technology Program(2024NSFTD0049,2024ZDZX0004).

PDF (PC)

76

摘要： 掩码图自编码器(Masked Graph Autoencoders,MGAEs)因能够有效处理图结构数据的节点分类任务而受到广泛关注。现有的掩码图自编码器模型在预训练编码器的过程中,存在语义信息损失和掩码节点嵌入相似两方面的不足。针对上述问题,提出一种融合对比学习的掩码图自编码器模型(CMGAE)。首先,将掩码图和原图分别输入在线编码器和目标编码器,生成在线嵌入和目标嵌入。然后,通过信息补充模块将在线嵌入和目标嵌入进行相似度对比,补充损失的语义信息。同时,将在线嵌入输入辨别函数和解码器,前者适当扩大掩码节点嵌入之间的方差,缓解掩码节点嵌入相似的问题,后者得到重构节点特征,用于训练在线编码器。最后,将预训练结束的在线编码器用于节点分类任务。在5个转导公共数据集和1个归纳数据集上进行节点分类实验,CMGAE的转导数据集准确率分别达到85.0%,73.6%,60.0%,50.5%,71.8%,归纳数据集的Micro-F1分数达到74.8%,相较于现有模型有着更好的性能。

关键词: 图神经网络, 节点分类, 掩码图自编码器, 图自监督学习, 图对比学习

Abstract: MGAEs have gained significant attention due to their effectiveness in handling node classification tasks on graph-structured data.However,existing MGAE models face two main limitations during the pretraining of the encoder:semantic information loss,and similarity of embeddings for masked nodes.To mitigate these issues,this paper proposes a Contrastive Masked Graph Autoencoder model(CMGAE).Firstly,the masked graph and the original graph are separately fed into the online encoder and the target encoder to generate online embeddings and target embeddings,respectively.Then,an information supplementation module is employed to compare the similarity between the online embeddings and target embeddings,thereby recovering the lost semantic information.Simultaneously,the online embeddings are passed through a discriminator function and decoder.The discriminator function helps increase the variance of the embeddings for masked nodes,mitigating the issue of similar embeddings for masked nodes.The decoder reconstructs node features that are used to train the online encoder.Finally,the pretrained online encoder is utilized for node classification tasks.Node classification experiments are conducted on five transductive benchmark datasets and one inductive dataset.The results show that CMGAE achieves a transductive accuracy of 85.0%,73.6%,60.0%,50.5%,and 71.8% on the respective datasets,while the Micro-F1 score on the inductive dataset reaches 74.8%.These results demonstrate that CMGAE outperforms existing models.

Key words: Graph neural network, Node classification, Masked graph autoencoder, Graph self-supervised learning, Graph contrastive learning

中图分类号: 

  • TP391
[1]WANG K,SHEN Z,HUANG C,et al.Microsoft academicgraph:When experts are not enough[J].Quantitative Science Studies,2020,1(1):396-413.
[2]MERNYEI P,CANGEA C.Wiki-cs:A wikipedia-based bench-mark for graph neural networks[J].arXiv:2007.02901,2020.
[3]FOUT A,BYRD J,SHARIAT B,et al.Protein interface prediction using graph convolutional networks[C]//Advances in Neural Information Processing Systems.2017.
[4]ZHANG L Y,SUN H H,SUN Y F,et al.Review of Node Classification Methods Based on Graph Convolutional Neural Networks[J].Computer Science,2024,51(4):95-105.
[5]LI X,LU W,MA Z Y,et al.A Node Classification MethodBased on Graph Attention and Improved Transformer[J].Chinese Journal of Electronics,2024,52(8):2799-2810.
[6]LIU Y L,QIU R H,TANG Y R,et al.PUMA:Efficient Conti-nual Graph Learning for Node Classification With Graph Condensation[J].IEEE Transactions on Knowledge and Data Engineering,2025,37(1):449-461.
[7]DUAN L,CHEN X,LIU W,et al.Structural entropy basedgraph structure learning for node classification[C]//Procee-dings of the AAAI Conference on Artificial Intelligence.2024:8372-8379.
[8]KIPF T N,WELLING M.Semi-supervised classification withgraph convolutional networks[J].arXiv:1609.02907,2016.
[9]VELIČKOVIĆ P,CUCURULL G,CASANOVA A,et al.Graph attention networks[J].arXiv:1710.10903,2017.
[10]XU K,HU W,LESKOVEC J,et al.How powerful are graphneural networks?[J].arXiv:1810.00826,2018.
[11]HE K,CHEN X,XIE S,et al.Masked autoencoders are scalable vision learners[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:16000-16009.
[12]HOU Z,LIU X,CEN Y,et al.GraphMAE:Self-supervisedmasked graph autoencoders[C]//Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mi-ning.2022:594-604.
[13]HOU Z,HE Y,CEN Y,et al.GraphMAE2:A decoding-enhanced masked self-supervised graph learner[C]//Proceedings of the ACM Web Conference 2023.2023:737-746.
[14]TU W X,LIAO Q,ZHOU S H,et al.Rare:Robust masked graph autoencoder[J].IEEE Transactions on Knowledge and Data Engineering,2024,36(10):5340-5353.
[15]WANG L,TAO X,LIU Q,et al.Rethinking Graph Masked Autoencoders through Alignment and Uniformity[C]//Procee-dings of the AAAI Conference on Artificial Intelligence.2024:15528-15536.
[16]LI J,WU R,SUN W,et al.What’s Behind the Mask:Understanding Masked Graph Modeling for Graph Autoencoders[C]//Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.2023:1268-1279.
[17]PEROZZI B,AL-RFOU R,SKIENA S.Deepwalk:Online lear-ning of social representations[C]//Proceedings of the 20th ACM SIGKDD International Conference on Knowledge Disco-very and Data Mining.2014:701-710.
[18]TAN Q,LIU N,HUANG X,et al.S2gae:Self-supervised graph autoencoders are generalizable learners with graph masking[C]//Proceedings of the Sixteenth ACM International Confe-rence on Web Search and Data Mining.2023:787-795.
[19]ZHAO Z,LI Y,ZOU Y,et al.Masked Graph Autoencoder with Non-discreteBandwidths[C]//Proceedings of the ACM on Web Conference.2024.2024:377-388.
[20]LI X,YE T,SHAN C,et al.Seegera:Self-supervised semi-implicit graph variational auto-encoders with masking[C]//Proceedings of the ACM Web Conference 2023.2023:143-153.
[21]VELIČKOVIĆ P,FEDUS W,HAMILTON W L,et al.Deep graph infomax[J].arXiv:1809.10341,2018.
[22]HASSANI K,KHASAHMADI A H.Contrastive multi-viewrepresentation learning on graphs[C]//International Confe-rence on Machine Learning.PMLR,2020:4116-4126.
[23]ZHANG H,WU Q,YAN J,et al.From canonical correlationanalysis to self-supervised graph neural networks[J].Advances in Neural Information Processing Systems,2021,34:76-89.
[24]ZHU Y,XU Y,YU F,et al.Deep graph contrastive representation learning[J].arXiv:2006.04131,2020.
[25]ZHU Y,XU Y,YU F,et al.Graph contrastive learning withadaptive augmentation[C]//Proceedings of the Web Conference 2021.2021:2069-2080.
[26]THAKOOR S,TALLEC C,AZAR M G,et al.Large-scale representation learning on graphs via bootstrapping[J].arXiv:2102.06514,2021.
[27]JI C,HUANG Z,SUN Q,et al.ReGCL:rethinking messagepassing in graph contrastive learning[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2024:8544-8552.
[28]WU C,WANG C,XU J,et al.Graph contrastive learning with generative adversarial network[C]//Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining.2023:2721-2730.
[29]SEN P,NAMATA G,BILGIC M,et al.Collective classification in network data[J].AI Magazine,2008,29(3):93-93.
[30]BOJCHEVSKI A,GÜNNEMANN S.Deep Gaussian embedding of graphs:Unsupervised inductive learning via ranking[J].ar-Xiv:1707.03815,2017.
[31]ZENG H,ZHOU H,SRIVASTAVA A,et al.Graphsaint:Graph sampling based inductive learning method[J].arXiv:1907.04931,2019.
[32]HU W H,FEY M,ZITNIK M,et al.Open Graph Benchmark:Datasets for Machine Learning on Graphs[C]//Proceedings of the 34th International Conference on Neural Information Processing Systems.Red Hook,NY:Curran Associates Inc.,2020:22118-22133.
[33]HAMILTON W,YING Z,LESKOVEC J.Inductive representa-tion learning on large graphs[C]//Advances in Neural Information Processing Systems.2017.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
渝ICP备16013121号-4
地址：重庆市两江新区洪湖西路18号    邮编：401121  
电话：023-63500828（编辑部） 023-67039612（会议/专题） 023-67039623（财务/发票）
E-mail：jsjkx12@163.com
本系统由北京玛格泰克科技发展有限公司设计开发