Phase transition of Surprise optimization in community detection
Phase transition of Surprise optimization in community detection
复制标题
社区检测中惊喜优化的相变
DOI：
10.1016/j.physa.2017.09.090
复制
发表时间：
2018-02
期刊：
影响因子：
--
通讯作者：
Chen Shi
中科院分区：
文献类型：
--
作者：
Xiang Ju;Tang Yan-Ni;Gao Yuan-Yuan;Liu Lang;Hao Yi;Li Jian-Ming;Zhang Yan;Chen Shi
社区检测是复杂网络研究中的重要问题之一。在文献中，提出了许多方法来检测网络中的社区结构，同时它们本身也有适用范围。本文研究了一种重要的群落检测方法——Surprise （Aldecoa）和Marín， Sci。众议员3(2013)1060)，通过关注社区合并和分裂的关键点。首先分析了群落划分跃迁的临界行为，给出了群落划分跃迁的相图。结果表明：惊喜群落的临界数量随连接密度差的增加呈超指数增长，而群落间和群落内连接密度差异较小，其临界数量接近模块化的临界数量；通过直接优化Surprise，我们在各种网络上对结果进行了实验测试，并与其他经典方法进行了一系列比较，进一步发现网络的异质性会加速社区的分裂。总体而言，由于链路密度、程度和社区规模的异质性等多种原因，Surprise倾向于分裂社区，因此在社区检测中表现出比Modularity等其他方法更高的分辨率。最后，我们提供了几种增强Surprise的方法。
Community detection is one of important issues in the research of complex networks. In literatures, many methods have been proposed to detect community structures in the networks, while they also have the scope of application themselves. In this paper, we investigate an important measure for community detection, Surprise (Aldecoa and Marín, Sci. Rep. 3 (2013) 1060), by focusing on the critical points in the merging and splitting of communities. We firstly analyze the critical behavior of Surprise and give the phase diagrams in community-partition transition. The results show that the critical number of communities for Surprise has a super-exponential increase with the increase of the link-density difference, while it is close to that of Modularity for small difference between inter- and intra-community link densities. By directly optimizing Surprise, we experimentally test the results on various networks, following a series of comparisons with other classical methods, and further find that the heterogeneity of networks could quicken the splitting of communities. On the whole, the results show that Surprise tends to split communities due to various reasons such as the heterogeneity in link density, degree and community size, and it thus exhibits higher resolution than other methods, e.g., Modularity, in community detection. Finally, we provide several approaches for enhancing Surprise.
登录
查看更多内容
影响因子：
1.9
作者：
Jinlong Ma;Weizhan Han;Qing Guo;Shuai Zhang;Junfang Wang;Zhihao Wang
通讯作者：
Jinlong Ma;Weizhan Han;Qing Guo;Shuai Zhang;Junfang Wang;Zhihao Wang
DOI：
10.1016/j.physa.2016.11.019
发表时间：
2017
期刊：
Physica A:Statistical Mechanics and Its Applications
影响因子：
--
作者：
He Jialin;Chen Duanbing;Sun Chongjing;Fu Yan;Li Wenjun
通讯作者：
Li Wenjun
利用二分网络中检测到的社区改进推荐算法
DOI：
10.1016/j.physa.2016.11.076
发表时间：
2017-04
期刊：
Physica A
影响因子：
--
作者：
Peng Zhang;Duo Wang;Jinghua Xiao
通讯作者：
Jinghua Xiao
通过最大化模块化密度来确定蛋白质相互作用网络的模块化组织
影响因子：
--
作者：
Zhang S;Ning XM;Ding C;Zhang XS
通讯作者：
Zhang XS
DOI：
10.1364/ofc.2016.w2a.51
发表时间：
2016-03
期刊：
2016 Optical Fiber Communications Conference and Exhibition (OFC)
影响因子：
--
作者：
Haoni Li;Nan Wang;Ping Gong;Edward J. Perkins;Chaoyang Zhang
通讯作者：
Haoni Li;Nan Wang;Ping Gong;Edward J. Perkins;Chaoyang Zhang