An evolutionary autoencoder for dynamic community detection
- PDF / 1,459,655 Bytes
- 16 Pages / 595 x 842 pts (A4) Page_size
- 84 Downloads / 245 Views
. RESEARCH PAPER .
November 2020, Vol. 63 212205:1–212205:16 https://doi.org/10.1007/s11432-020-2827-9
An evolutionary autoencoder for dynamic community detection Zhen WANG1 , Chunyu WANG2 , Chao GAO2,3 , Xuelong LI3* & Xianghua LI3,2* 1
School of Mechanical Engineering and The Center for OPTIMAL, Northwestern Polytechnical University, Xi’an 710072, China; 2 College of Computer and Information Science, Southwest University, Chongqing 400715, China; 3 School of Computer Science and The Center for OPTIMAL, Northwestern Polytechnical University, Xi’an 710072, China Received 15 February 2020/Accepted 12 March 2020/Published online 28 September 2020
Abstract Dynamic community detection is significant for controlling and capturing the temporal features of networks. The evolutionary clustering framework provides a temporal smoothness constraint for simultaneously maximizing the clustering quality at the current time step and minimizing the clustering deviation between two successive time steps. Based on this framework, some existing methods, such as the evolutionary spectral clustering and evolutionary nonnegative matrix factorization, aim to look for the low-dimensional representation by mapping reconstruction. However, such reconstruction does not address the nonlinear characteristics of networks. In this paper, we propose a semi-supervised algorithm (sE-Autoencoder) to overcome the effects of nonlinear property on the low-dimensional representation. Our proposed method extends the typical nonlinear reconstruction model to the dynamic network by constructing a temporal matrix. More specifically, the potential community characteristics and the previous clustering, as the prior information, are incorporated into the loss function as a regularization term. Experimental results on synthetic and realworld datasets demonstrate that the proposed method is effective and superior to other methods for dynamic community detection. Keywords
dynamic networks, community detection, autoencoder, graph embedding
Citation Wang Z, Wang C Y, Gao C, et al. An evolutionary autoencoder for dynamic community detection. Sci China Inf Sci, 2020, 63(11): 212205, https://doi.org/10.1007/s11432-020-2827-9
1
Introduction
Many real-world systems can be formulated as complex networks (e.g., scientific collaboration networks [1], power networks [2], biological networks [3], traffic networks [4, 5] and disease networks [6]). The structure analysis of complex networks is a vigoroso and easy-to-implement tool to comprehend and forecast the functions and features of a network [7, 8]. The community structure, as one of the most crucial and essential structure characteristics, acts on the dynamic trait of a network [9]. It can be abstracted into the aggregation of nodes with the sparse inter-community links and dense intra-community links [10]. In reality, many networks derived from real-world systems have the dynamic characteristics, but most existing algorithms are centered on detecting community structure in static networks [11]. For instance, in
Data Loading...
