Paper Info
Title
Incomplete multi-view clustering based on weighted sparse and low rank representation
Abstract
Multi-view clustering utilizes the consistency and complementarity between views to group entities well. However, in real life, the lack of instances in some views often occurs, which not only reduces the available information, but also increases the difficulty of joint learning with non-aligned multi-view data. Many incomplete multi-view clustering algorithms are developed to tackle these concerns, but they usually have the following problems: 1) They mainly focus on how to construct the shared feature space for incomplete views while ignoring the essential relationship between data instances. 2) Most of them simply assume that two datapoints which are close belong to the same category, but that is not the case. 3) The hazards of overlapping, confusing features in incomplete multi-view clustering are not considered. To solve these issues, this paper proposes a new Incomplete Multi-view Graph Learning method based on Weighted Sparse and Low rank Representation (IMGLWSLR). It leverages subspace learning with double constraints to capture global and local data relationships, a weighting mechanism to reduce the negative impact of missing data and a kernel-based method to fuse incomplete multiple views. Different from previous approaches, we concentrate on inhibiting the confusion of redundant features in subspace learning, which may affect the clustering seriously with missing views. Experimental results demonstrate the superiority of IMGLWSLR over nine benchmark datasets, compared with seven state-of-the-art approaches.
Year
DOI
Venue
2022
10.1007/s10489-022-03246-4
Applied Intelligence
Keywords
DocType
Volume
Incomplete multi-view clustering, Graph learning, Sparse and low rank representation, Weighting mechanism
Journal
52
Issue
ISSN
Citations 
13
0924-669X
0
PageRank 
References 
Authors
0.34
19
4
Name
Order
Citations
PageRank
Zhao Liang100.34
Jie Zhang221430.67
Tao Yang358.53
Zhikui Chen469266.76