HCL: Improving Graph Representation with Hierarchical Contrastive Learning

被引:0
作者
Wang, Jun [1 ]
Li, Weixun [1 ]
Hou, Changyu [1 ]
Tang, Xin [2 ]
Qiao, Yixuan [1 ]
Fang, Rui [2 ]
Li, Pengyong [3 ]
Gao, Peng [1 ]
Xie, Guotong [1 ,4 ,5 ]
机构
[1] Ping Healthcare Technol, Beijing, Peoples R China
[2] Ping Property & Casualty Insurance Co, Shenzhen, Peoples R China
[3] Xidian Univ, Sch Comp Sci & Technol, Xian, Peoples R China
[4] Ping Hlth Cloud Co Ltd, Shenzhen, Peoples R China
[5] Ping Int Smart City Technol Co Ltd, Shenzhen, Peoples R China
来源
SEMANTIC WEB - ISWC 2022 | 2022年 / 13489卷
关键词
Data mining; Graph learning; Contrastive learning;
D O I
10.1007/978-3-031-19433-7_7
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Contrastive learning has emerged as a powerful tool for graph representation learning. However, most contrastive learning methods learn features of graphs with fixed coarse-grained scale, which might underestimate either local or global information. To capture more hierarchical and richer representation, we propose a novel Hierarchical Contrastive Learning (HCL) framework that explicitly learns graph representation in a hierarchical manner. Specifically, HCL includes two key components: a novel adaptive Learning to Pool (L2Pool) method to construct more reasonable multi-scale graph topology for more comprehensive contrastive objective, a novel multi-channel pseudo-siamese network to further enable more expressive learning of mutual information within each scale. Comprehensive experimental results show HCL achieves competitive performance on 12 datasets involving node classification, node clustering and graph classification. In addition, the visualization of learned representation reveals that HCL successfully captures meaningful characteristics of graphs.
引用
收藏
页码:108 / 124
页数:17
相关论文
共 48 条
  • [1] Sub2Vec: Feature Learning for Subgraphs
    Adhikari, Bijaya
    Zhang, Yao
    Ramakrishnan, Naren
    Prakash, B. Aditya
    [J]. ADVANCES IN KNOWLEDGE DISCOVERY AND DATA MINING, PAKDD 2018, PT II, 2018, 10938 : 170 - 182
  • [2] Shortest-path kernels on graphs
    Borgwardt, KM
    Kriegel, HP
    [J]. Fifth IEEE International Conference on Data Mining, Proceedings, 2005, : 74 - 81
  • [3] ChemDB update - full-text search and virtual chemical space
    Chen, Jonathan H.
    Linstead, Erik
    Swamidass, S. Joshua
    Wang, Dennis
    Baldi, Pierre
    [J]. BIOINFORMATICS, 2007, 23 (17) : 2348 - 2351
  • [4] Chen Ming., 2020, ICML
  • [5] Exploring Simple Siamese Representation Learning
    Chen, Xinlei
    He, Kaiming
    [J]. 2021 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR 2021, 2021, : 15745 - 15753
  • [6] Defferrard M, 2016, ADV NEUR IN, V29
  • [7] Fedus W. L., 2019, P ICLR, P1
  • [8] Gao HY, 2019, PR MACH LEARN RES, V97
  • [9] On graph kernels:: Hardness results and efficient alternatives
    Gärtner, T
    Flach, P
    Wrobel, S
    [J]. LEARNING THEORY AND KERNEL MACHINES, 2003, 2777 : 129 - 143
  • [10] Gilmer J, 2017, PR MACH LEARN RES, V70