Co-consensus semi-supervised multi-view learning with orthogonal non-negative matrix factorization
作者：
Highlights：
• Different from the existing semi-supervised multi-view learning methods, the proposed method can exploit both the consensus relations between samples and between samples and their assemble centroid.
• To simultaneously implement the orthogonality and the label constraints, we fix the label information in the representation and design a partial orthogonality constraint, which will lead to full orthogonality constraint. Additionally, we provide a proposition about the equivalency of these two orthogonality constraints.
• To solve the proposed model, we design an effective iterative algorithm, and theoretically provide its analysis on the convergence.
摘要
•Different from the existing semi-supervised multi-view learning methods, the proposed method can exploit both the consensus relations between samples and between samples and their assemble centroid.•To simultaneously implement the orthogonality and the label constraints, we fix the label information in the representation and design a partial orthogonality constraint, which will lead to full orthogonality constraint. Additionally, we provide a proposition about the equivalency of these two orthogonality constraints.•To solve the proposed model, we design an effective iterative algorithm, and theoretically provide its analysis on the convergence.
论文关键词：Multi-view learning,Semi-supervised learning,Orthogonal constraint,Orthogonal non-negative matrix factorization,Consensus
论文评审过程：Received 17 July 2022, Revised 30 July 2022, Accepted 30 July 2022, Available online 19 August 2022, Version of Record 19 August 2022.
论文官网地址：https://doi.org/10.1016/j.ipm.2022.103054
