Multi-view clustering by non-negative matrix factorization with co-orthogonal constraints

Non-negative matrix factorization (NMF) has attracted sustaining attention in multi-view clustering, because of its ability of processing high-dimensional data. In order to learn the desired dimensional-reduced representation, a natural scheme is to add constraints to traditional NMF. Motivated by t...

Full description

Saved in:
Bibliographic Details
Published in:Knowledge-based systems 2020-04, Vol.194, p.105582, Article 105582
Main Authors: Liang, Naiyao, Yang, Zuyuan, Li, Zhenni, Sun, Weijun, Xie, Shengli
Format: Article
Language:English
Subjects:
Algorithms
Clustering
Co-orthogonal constraints
Constraint modelling
Factorization
Iterative methods
Mathematical analysis
Matrix methods
Multi-view clustering
Non-negative matrix factorization
Orthogonality
Representations
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Non-negative matrix factorization (NMF) has attracted sustaining attention in multi-view clustering, because of its ability of processing high-dimensional data. In order to learn the desired dimensional-reduced representation, a natural scheme is to add constraints to traditional NMF. Motivated by that the clustering performance is affected by the orthogonality of inner vectors of both the learned basis matrices and the representation matrices, a novel NMF model with co-orthogonal constraints is designed to deal with the multi-view clustering problem in this paper. For solving the proposed model, an efficient iterative updating algorithm is derived. And the corresponding convergence is proved, together with the analysis to its computational complexity. Experiments on five datasets are performed to present the advantages of the proposed algorithm against the state-of-the-art methods.
ISSN:0950-7051
1872-7409
DOI:10.1016/j.knosys.2020.105582