How Robust Are Reconstruction Thresholds for Community Detection?

被引:38
|
作者
Moitra, Ankur [1 ]
Perry, William [1 ]
Wein, Alexander S. [1 ]
机构
[1] MIT, 77 Massachusetts Ave, Cambridge, MA 02139 USA
来源
STOC'16: PROCEEDINGS OF THE 48TH ANNUAL ACM SIGACT SYMPOSIUM ON THEORY OF COMPUTING | 2016年
关键词
Community detection; stochastic block model; semirandom models; broadcast tree model; semidefinite programming;
D O I
10.1145/2897518.2897573
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
The stochastic block model is one of the oldest and most ubiquitous models for studying clustering and community detection. In an exciting sequence of developments, motivated by deep but non-rigorous ideas from statistical physics, Decelle et al. conjectured a sharp threshold for when community detection is possible in the sparse regime. Mossel, Neeman and Sly and Massoulie proved the conjecture and gave matching algorithms and lower bounds. Here we revisit the stochastic block model from the perspective of semirandom models where we allow an adversary to make 'helpful' changes that strengthen ties within each community and break ties between them. We show a surprising result that these 'helpful' changes can shift the information-theoretic threshold, making the community detection problem strictly harder. We complement this by showing that an algorithm based on semidefinite programming (which was known to get close to the threshold) continues to work in the semirandom model (even for partial recovery). This suggests that algorithms based on semidefinite programming are robust in ways that any algorithm meeting the information-theoretic threshold cannot be. These results point to an interesting new direction: Can we find robust, semirandom analogues to some of the classical, average-case thresholds in statistics? We also explore this question in the broadcast tree model, and we show that the viewpoint of semirandom models can help explain why some algorithms are preferred to others in practice, in spite of the gaps in their statistical performance on random models.
引用
收藏
页码:828 / 841
页数:14
相关论文
共 50 条
  • [21] An optimisation tool for robust community detection algorithms using content and topology information
    Amhmed Bhih
    Princy Johnson
    Martin Randles
    The Journal of Supercomputing, 2020, 76 : 226 - 254
  • [22] Joint Community Detection and Rotational Synchronization via Semidefinite Programming
    Fan, Yifeng
    Khoo, Yuehaw
    Zhao, Zhizhen
    SIAM JOURNAL ON MATHEMATICS OF DATA SCIENCE, 2022, 4 (03): : 1052 - 1081
  • [23] Community detection with nodal information: Likelihood and its variational approximation
    Weng, Haolei
    Feng, Yang
    STAT, 2022, 11 (01):
  • [24] GBTM: Community detection and network reconstruction for noisy and time-evolving data
    Chen, Xiao
    Hu, Jie
    Chen, Yu
    INFORMATION SCIENCES, 2024, 679
  • [25] GATFELPA integrates graph attention networks and enhanced label propagation for robust community detection
    Tang, Feiyi
    Li, Junxian
    Liu, Xi
    Chang, Chao
    Teng, Luyao
    SCIENTIFIC REPORTS, 2025, 15 (01):
  • [26] Community detection method based on robust semi-supervised nonnegative matrix factorization
    He, Chaobo
    Zhang, Qiong
    Tang, Yong
    Liu, Shuangyin
    Zheng, Jianhua
    PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS, 2019, 523 : 279 - 291
  • [27] Is it correct to project and detect? How weighting unipartite projections influences community detection
    Cann, Tristan J. B.
    Weaver, Iain S.
    Williams, Hywel T. P.
    NETWORK SCIENCE, 2020, 8 : S145 - S163
  • [28] Community detection with a subsampled semidefinite program
    Abdalla, Pedro
    Bandeira, Afonso S.
    SAMPLING THEORY SIGNAL PROCESSING AND DATA ANALYSIS, 2022, 20 (01):
  • [29] Community detection with structural and attribute similarities
    Tang, Fengqin
    Ding, Wenwen
    JOURNAL OF STATISTICAL COMPUTATION AND SIMULATION, 2019, 89 (04) : 668 - 685
  • [30] Dynamic network sampling for community detection
    Mu, Cong
    Park, Youngser
    Priebe, Carey E. E.
    APPLIED NETWORK SCIENCE, 2023, 8 (01)