Graph Communal Contrastive Learning

被引:48
作者
Li, Bolian [1 ]
Jing, Baoyu [2 ]
Tong, Hanghang [2 ]
机构
[1] Tianjin Univ, Tianjin, Peoples R China
[2] Univ Illinois, Champaign, IL USA
来源
PROCEEDINGS OF THE ACM WEB CONFERENCE 2022 (WWW'22) | 2022年
基金
美国国家科学基金会;
关键词
self-supervised learning; graph contrastive learning; community detection; NETWORK;
D O I
10.1145/3485447.3512208
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
Graph representation learning is crucial for many real-world applications (e.g. social relation analysis). A fundamental problem for graph representation learning is how to effectively learn representations without human labeling, which is usually costly and time-consuming. Graph contrastive learning (GCL) addresses this problem by pulling the positive node pairs (or similar nodes) closer while pushing the negative node pairs (or dissimilar nodes) apart in the representation space. Despite the success of the existing GCL methods, they primarily sample node pairs based on the node-level proximity yet the community structures have rarely been taken into consideration. As a result, two nodes from the same community might be sampled as a negative pair. We argue that the community information should be considered to identify node pairs in the same communities, where the nodes insides are semantically similar. To address this issue, we propose a novel Graph Communal Contrastive Learning (gCooL) framework to jointly learn the community partition and learn node representations in an end-to-end fashion. Specifically, the proposed gCooL consists of two components: a Dense Community Aggregation (DeCA) algorithm for community detection and a Reweighted Self-supervised Cross-contrastive (ReSC) training scheme to utilize the community information. Additionally, the real-world graphs are complex and often consist of multiple views. In this paper, we demonstrate that the proposed gCooL can also be naturally adapted to multiplex graphs. Finally, we comprehensively evaluate the proposed gCooL on a variety of real-world graphs. The experimental results show that the gCooL outperforms the state-of-the-art methods.
引用
收藏
页码:1203 / 1213
页数:11
相关论文
共 78 条
[1]  
[Anonymous], 2021, DYDIFF VAE DYNAMIC V, DOI DOI 10.1145/3404835.3462934
[2]  
[Anonymous], 2018, ICLR, DOI DOI 10.1109/ISCAS.2018.8350934
[3]   SELF-ORGANIZING NEURAL NETWORK THAT DISCOVERS SURFACES IN RANDOM-DOT STEREOGRAMS [J].
BECKER, S ;
HINTON, GE .
NATURE, 1992, 355 (6356) :161-163
[4]  
Bhagat S, 2011, SOCIAL NETWORK DATA ANALYTICS, P115
[5]   Normalized Gaussian radial basis function networks [J].
Bugmann, G .
NEUROCOMPUTING, 1998, 20 (1-3) :97-110
[6]  
Caron Mathilde, 2020, 24 C NEUR INF PROC S
[7]   Representation Learning for Attributed Multiplex Heterogeneous Network [J].
Cen, Yukuo ;
Zou, Xu ;
Zhang, Jianwei ;
Yang, Hongxia ;
Zhou, Jingren ;
Tang, Jie .
KDD'19: PROCEEDINGS OF THE 25TH ACM SIGKDD INTERNATIONAL CONFERENCCE ON KNOWLEDGE DISCOVERY AND DATA MINING, 2019, :1358-1368
[8]   Heterogeneous Network Embedding via Deep Architectures [J].
Chang, Shiyu ;
Han, Wei ;
Tang, Jiliang ;
Qi, Guo-Jun ;
Aggarwal, Charu C. ;
Huang, Thomas S. .
KDD'15: PROCEEDINGS OF THE 21ST ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING, 2015, :119-128
[9]   Cross-Network Embedding for Multi-Network Alignment [J].
Chu, Xiaokai ;
Fan, Xinxin ;
Yao, Di ;
Zhu, Zhihua ;
Huang, Jianhui ;
Bi, Jingping .
WEB CONFERENCE 2019: PROCEEDINGS OF THE WORLD WIDE WEB CONFERENCE (WWW 2019), 2019, :273-284
[10]  
Chuang C.-Y., 2020, NeurIPS, DOI DOI 10.48550/ARXIV.2007.00224