Optimization of identifiability for efficient community detection

被引:124
作者
Li, Hui-Jia [1 ]
Wang, Lin [2 ]
Zhang, Yan [3 ]
Perc, Matjaz [4 ,5 ,6 ]
机构
[1] Beijing Univ Posts & Telecommun, Sch Sci, Beijing 100876, Peoples R China
[2] Univ Cambridge, Dept Genet, Cambridge CB2 3EH, England
[3] Alibaba Grp, Alibaba Local Serv Lab, Shanghai 200333, Peoples R China
[4] Univ Maribor, Fac Nat Sci & Math, Koroska Cesta 160, Maribor 2000, Slovenia
[5] China Med Univ, China Med Univ Hosp, Dept Med Res, Taichung, Taiwan
[6] Complex Sci Hub Vienna, Josefstadterstr 39, A-1080 Vienna, Austria
基金
中国国家自然科学基金; 北京市自然科学基金;
关键词
complex system; complex network; community detection; matrix factorization; COMPLEX NETWORKS; PHYSICS;
D O I
10.1088/1367-2630/ab8e5e
中图分类号
O4 [物理学];
学科分类号
0702 ;
摘要
Many physical and social systems are best described by networks. And the structural properties of these networks often critically determine the properties and function of the resulting mathematical models. An important method to infer the correlations between topology and function is the detection of community structure, which plays a key role in the analysis, design, and optimization of many complex systems. The nonnegative matrix factorization has been used prolifically to that effect in recent years, although it cannot guarantee balanced partitions, and it also does not allow a proactive computation of the number of communities in a network. This indicates that the nonnegative matrix factorization does not satisfy all the nonnegative low-rank approximation conditions. Here we show how to resolve this important open problem by optimizing the identifiability of community structure. We propose a new form of nonnegative matrix decomposition and a probabilistic surrogate learning function that can be solved according to the majorization-minimization principle. Extensivein silicotests on artificial and real-world data demonstrate the efficient performance in community detection, regardless of the size and complexity of the network.
引用
收藏
页数:10
相关论文
共 49 条
[21]   Community detection algorithms: A comparative analysis [J].
Lancichinetti, Andrea ;
Fortunato, Santo .
PHYSICAL REVIEW E, 2009, 80 (05)
[22]   Benchmark graphs for testing community detection algorithms [J].
Lancichinetti, Andrea ;
Fortunato, Santo ;
Radicchi, Filippo .
PHYSICAL REVIEW E, 2008, 78 (04)
[23]   The science of fake news [J].
Lazer, David M. J. ;
Baum, Matthew A. ;
Benkler, Yochai ;
Berinsky, Adam J. ;
Greenhill, Kelly M. ;
Menczer, Filippo ;
Metzger, Miriam J. ;
Nyhan, Brendan ;
Pennycook, Gordon ;
Rothschild, David ;
Schudson, Michael ;
Sloman, Steven A. ;
Sunstein, Cass R. ;
Thorson, Emily A. ;
Watts, Duncan J. ;
Zittrain, Jonathan L. .
SCIENCE, 2018, 359 (6380) :1094-1096
[24]   Learning the parts of objects by non-negative matrix factorization [J].
Lee, DD ;
Seung, HS .
NATURE, 1999, 401 (6755) :788-791
[25]   Measurability of the epidemic reproduction number in data-driven contact networks [J].
Liu, Quan-Hui ;
Ajelli, Marco ;
Aleta, Alberto ;
Merler, Stefano ;
Moreno, Yamir ;
Vespignani, Alessandro .
PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA, 2018, 115 (50) :12680-12685
[26]   The physics of brain network structure, function and control [J].
Lynn, Christopher W. ;
Bassett, Danielle S. .
NATURE REVIEWS PHYSICS, 2019, 1 (05) :318-332
[27]  
Mahajan M, 2009, LECT NOTES COMPUT SC, V5431, P274, DOI 10.1007/978-3-642-00202-1_24
[28]   The structure and function of complex networks [J].
Newman, MEJ .
SIAM REVIEW, 2003, 45 (02) :167-256
[29]   Quantifying randomness in real networks [J].
Orsini, Chiara ;
Dankulov, Marija M. ;
Colomer-de-Simon, Pol ;
Jamakovic, Almerima ;
Mahadevan, Priya ;
Vahdat, Amin ;
Bassler, Kevin E. ;
Toroczkai, Zoltan ;
Boguna, Marian ;
Caldarelli, Guido ;
Fortunato, Santo ;
Krioukov, Dmitri .
NATURE COMMUNICATIONS, 2015, 6
[30]   Statistical physics of human cooperation [J].
Perc, Matjaz ;
Jordan, Jillian J. ;
Rand, David G. ;
Wang, Zhen ;
Boccaletti, Stefano ;
Szolnoki, Attila .
PHYSICS REPORTS-REVIEW SECTION OF PHYSICS LETTERS, 2017, 687 :1-51