Z-Laplacian Matrix Factorization: Network Embedding With Interpretable Graph Signals

被引:0
作者
Wan, Liangtian [1 ,2 ]
Fu, Zhengqiang [1 ,2 ]
Ling, Yi [3 ]
Sun, Yuchen [1 ,2 ]
Li, Xiaona [1 ,2 ]
Sun, Lu [4 ]
Xia, Feng [5 ]
Yan, Xiaoran [6 ]
Aggarwal, Charu C. [7 ]
机构
[1] Dalian Univ Technol, DUT Sch Software Technol, Key Lab Ubiquitous Network & Serv Software Liaonin, Dalian 116620, Peoples R China
[2] Dalian Univ Technol, DUT RU Int Sch Informat Sci & Engn, Dalian 116620, Peoples R China
[3] Amazon, Atlanta, GA 30337 USA
[4] Dalian Maritime Univ, Inst Informat Sci Technol, Dept Commun Engn, Dalian 116026, Peoples R China
[5] RMIT Univ, Sch Comp Technol, Melbourne, Vic 3000, Australia
[6] Zhejiang Lab, Res Inst Artificial Intelligence, Hangzhou 311121, Peoples R China
[7] IBM T J Watson Res Ctr, Hawthorne, NY 10598 USA
基金
国家重点研发计划; 中国国家自然科学基金;
关键词
Matrix decomposition; Signal processing algorithms; Task analysis; Information filters; Electronic mail; Sun; Sparse matrices; Biased random walk; graph laplacian; link prediction; matrix factorization; network embedding; node classification;
D O I
10.1109/TKDE.2023.3331027
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Network embedding aims to represent nodes with low dimensional vectors while preserving structural information. It has been recently shown that many popular network embedding methods can be transformed into matrix factorization problems. In this paper, we propose the unifying framework "Z-NetMF," which generalizes random walk samplers to Z-Laplacian graph filters, leading to embedding algorithms with interpretable parameters. In particular, by controlling biases in the time domain, we propose the Z-NetMF-t algorithm, making it possible to scale contributions of random walks of different length. Inspired by node2vec, we design the Z-NetMF-g algorithm, capturing the random walk biases in the graph domain. Moreover, we evaluate the effect of the bias parameters based on node classification and link prediction tasks. The results show that our algorithms, especially the combined model Z-NetMF-gt with biases in both domains, outperform the state-of-art methods while providing interpretable insights at the same time. Finally, we discuss future directions of the Z-NetMF framework.
引用
收藏
页码:4331 / 4345
页数:15
相关论文
共 24 条
  • [21] Graph regularization weighted nonnegative matrix factorization for link prediction in weighted complex network
    Chen, Guangfu
    Xu, Chen
    Wang, Jingyi
    Feng, Jianwen
    Feng, Jiqiang
    NEUROCOMPUTING, 2019, 369 : 50 - 60
  • [22] GCNMFCDA: A Method Based on Graph Convolutional Network and Matrix Factorization for Predicting circRNA-Disease Associations
    Wang, Dian-Xiao
    Ji, Cun-Mei
    Wang, Yu-Tian
    Li, Lei
    Ni, Jian-Cheng
    Li, Bin
    INTELLIGENT COMPUTING THEORIES AND APPLICATION, ICIC 2022, PT II, 2022, 13394 : 166 - 180
  • [23] Multi-View Attributed Network Embedding Using Manifold Regularization Preserving Non-Negative Matrix Factorization
    Yuan, Weiwei
    Li, Xiang
    Guan, Donghai
    IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2024, 36 (06) : 2563 - 2571
  • [24] Multi-View Network Embedding Via Graph Factorization Clustering and Co-Regularized Multi-View Agreement
    Sun, Yiwei
    Bui, Ngot
    Hsieh, Tsung-Yu
    Honavar, Vasant
    2018 18TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS (ICDMW), 2018, : 1006 - 1013