Deep Adaptive Graph Clustering via von Mises-Fisher Distributions

doi:10.1145/3580521

	Deep Adaptive Graph Clustering via von Mises-Fisher Distributions
	Wang, Pengfei 1; Wu, Daqing 2; Chen, Chong 3; Liu, Kunpeng 5; Fu, Yanjie 4; Huang, Jianqiang 3; Zhou, Yuanchun 1; Zhan, Jianfeng 6; Hua, Xiansheng 3
	2024-05-01
发表期刊	ACM TRANSACTIONS ON THE WEB
ISSN	1559-1131
卷号	18 期号:2 页码:21
摘要	Graph clustering has been a hot research topic and is widely used in many fields, such as community detection in social networks. Lots of works combining auto-encoder and graph neural networks have been applied to clustering tasks by utilizing node attributes and graph structure. These works usually assumed the inherent parameters (i.e., size and variance) of different clusters in the latent embedding space are homogeneous, and hence the assigned probability is monotonous over the Euclidean distance between node embeddings and centroids. Unfortunately, this assumption usually does not hold since the size and concentration of different clusters can be quite different, which limits the clustering accuracy. In addition, the node embeddings in deep graph clustering methods are usually L2 normalized so that it lies on the surface of a unit hyper-sphere. To solve this problem, we proposed Deep Adaptive Graph Clustering via von Mises-Fisher distributions, namely DAGC. DAGC assumes the node embeddings H can be drawn from a von Mises-Fisher distribution and each cluster k is associated with cluster inherent parameters rho(k) which includes cluster center mu and cluster cohesion degree kappa. Then we adopt an EM-like approach (i.e., P(H\|rho) and P(rho\|H), respectively) to learn the embedding and cluster inherent parameters alternately. Specifically, with the node embeddings, we proposed to update the cluster centers in an attraction-repulsion manner to make the cluster centers more separable. And given the cluster inherent parameters, a likelihood-based loss is proposed to make node embeddings more concentrated around cluster centers. Thus, DAGC can simultaneously improve the intra-cluster compactness and inter-cluster heterogeneity. Finally, extensive experiments conducted on four benchmark datasets have demonstrated that the proposed DAGC consistently outperforms the state-of-the-art methods, especially on imbalanced datasets.
关键词	Graph embedding graph clustering vMF
DOI	10.1145/3580521
收录类别	SCI
语种	英语
资助项目	Natural Science Foundation of China[61836013] ; Strategic Priority Research Program of CAS[XDB31000000] ; Chinese Academy of Sciences Network Security and Informatization Special Application Demonstration Project[CAS-WX2021SF-0101-03]
WOS研究方向	Computer Science
WOS类目	Computer Science, Information Systems ; Computer Science, Software Engineering
WOS记录号	WOS:001208777200007
出版者	ASSOC COMPUTING MACHINERY
引用统计	被引频次：3[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/38993
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Fu, Yanjie
作者单位	1.Chinese Acad Sci, Comp Network Informat Ctr, CAS Informatization Plaza 2 Dong Sheng Nan Lu, Beijing 100083, Peoples R China 2.Peking Univ, DAMO Acad, Alibaba Grp, 969 West Wen Yi Rd, Hangzhou 311121, Peoples R China 3.DAMO Acad, Alibaba Grp, 969 West Wen Yi Rd, Hangzhou 311121, Peoples R China 4.Univ Cent Florida, 4000 Cent Florida Blvd, Orlando, FL 32816 USA 5.Portland State Univ, 1825 SW Broadway, Portland, OR 97201 USA 6.Chinese Acad Sci, Inst Comp Technol, 6 Kexueyuan South Rd Zhongguancun, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Wang, Pengfei,Wu, Daqing,Chen, Chong,et al. Deep Adaptive Graph Clustering via von Mises-Fisher Distributions[J]. ACM TRANSACTIONS ON THE WEB,2024,18(2):21.
APA	Wang, Pengfei.,Wu, Daqing.,Chen, Chong.,Liu, Kunpeng.,Fu, Yanjie.,...&Hua, Xiansheng.(2024).Deep Adaptive Graph Clustering via von Mises-Fisher Distributions.ACM TRANSACTIONS ON THE WEB,18(2),21.
MLA	Wang, Pengfei,et al."Deep Adaptive Graph Clustering via von Mises-Fisher Distributions".ACM TRANSACTIONS ON THE WEB 18.2(2024):21.