Penalized Flow Hypergraph Local Clustering

被引:0
作者
Zhong, Hao [1 ]
Zhang, Yubo [2 ]
Yan, Chenggang [1 ]
Xuan, Zuxing [4 ]
Yu, Ting [5 ]
Zhang, Ji [5 ]
Ying, Shihui [3 ]
Gao, Yue [2 ]
机构
[1] Hangzhou Dianzi Univ, Sch Automat, Hangzhou 310018, Peoples R China
[2] Tsinghua Univ, Sch Software, BNRist, KLISS,BLBCI,THUIBCS, Beijing 100190, Peoples R China
[3] Shanghai Univ, Sch Sci, Dept Math, Shanghai 200444, Peoples R China
[4] Beijing Union Univ, Inst Fundamental & Interdisciplinary Sci, Beijing 100101, Peoples R China
[5] Zhejiang Lab, Zhejiang 311100, Peoples R China
关键词
Community detection; hypergraph learning; local clustering; random walk;
D O I
10.1109/TKDE.2023.3319019
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In recent years, hypergraph analysis have attracted increasing attention due to their ability to model complex data correlation, with hypergraph clustering being one of the most important tasks. However, when the scale of hypergraph is large enough, clustering is difficult based on global consistency. Existing flow-based hypergraph local clustering methods have good theoretical cut improvements and runtime guarantees. However, these methods exhibit poor performance when the initial reference node set is small and are prone to causing the output set to shrink into a small subset, resulting in local minima. To address this issue, we propose the Penalized Flow Hypergraph Local Clustering(PFHLC) and provide new conductance guarantees and runtime analyses for our method. First, we use the random walk method to grow the initial seed set, and introduce the random walk information of nodes as penalized flow into the flow-based framework to optimize the output. Second, we propose a generalized objective function containing random walk information, which takes full advantage of the semi-supervised information of the target cluster to protect important nodes. This feature can avoid the local minima of previous flow-based methods. Importantly, our method is strongly-local and can run efficiently on large-scale hypergraphs. We contribute a real-world dataset and the experiments on real-world large-scale datasets show that PFHLC achieves the state-of-the-art significantly.
引用
收藏
页码:2110 / 2125
页数:16
相关论文
共 48 条
  • [1] Hypernetwork science via high-order hypergraph walks
    Aksoy, Sinan G.
    Joslyn, Cliff
    Marrero, Carlos Ortiz
    Praggastis, Brenda
    Purvine, Emilie
    [J]. EPJ DATA SCIENCE, 2020, 9 (01)
  • [2] Hypergraph Laplacians in Diffusion Framework
    Aktas, Mehmet Emin
    Akbas, Esra
    [J]. COMPLEX NETWORKS & THEIR APPLICATIONS X, VOL 2, 2022, 1016 : 277 - 288
  • [3] Clustering in graphs and hypergraphs with categorical edge labels
    Amburg, Ilya
    Veldt, Nate
    Benson, Austin
    [J]. WEB CONFERENCE 2020: PROCEEDINGS OF THE WORLD WIDE WEB CONFERENCE (WWW 2020), 2020, : 706 - 717
  • [4] Andersen R, 2006, ANN IEEE SYMP FOUND, P475
  • [5] Andersen R, 2008, PROCEEDINGS OF THE NINETEENTH ANNUAL ACM-SIAM SYMPOSIUM ON DISCRETE ALGORITHMS, P651
  • [6] [Anonymous], 1989, Hypergraphs: Combinatorics of finite sets
  • [7] [Anonymous], 2003, AAAI, DOI DOI 10.1145/2612669.2612699
  • [8] The Spacey Random Walk: A Stochastic Process for Higher-Order Data
    Benson, Austin R.
    Gleich, David F.
    Lim, Lek-Heng
    [J]. SIAM REVIEW, 2017, 59 (02) : 321 - 345
  • [9] Higher-order organization of complex networks
    Benson, Austin R.
    Gleich, David F.
    Leskovec, Jure
    [J]. SCIENCE, 2016, 353 (6295) : 163 - 166
  • [10] Blum A., 2001, P 18 INT C MACHINE L, P19, DOI [DOI 10.1184/R1/6606860.V1, 10.1184/R1/6606860.V1]