Subspace segmentation via self-regularized latent K-means

• Published in: Expert systems with applications 2019-12, Vol.136, p.316-326
• Main Authors: Wei, Lai, Zhou, Rigui, Wang, Xiaofeng, Zhu, Changming, Yin, Jun, Zhang, Xiafen
• Format: Article
• Language: English
• Subjects: Algorithms; Factorization; K-means; Low-rank representation; Matrix factorization; Optimization; Segmentation; Subspace segmentation; Subspaces
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2019.06.047

•A self-regularized latent K-means method was proposed for subspace segmentation.•The relationships between FRR-related algorithms SRLKM were built.•An optimization algorithm was proposed to solve SRLKM problem. Low-rank representation (LRR) related algorithms have achieved great successes in subspace segmentation tasks. Among them, matrix factorization based methods usually show better performances than those of other LRR-related algorithms. According to our analyses, we found that the matrix factorization based low-rank algorithms could be regarded as the extensions of classical K-means performed in a latent subspace. From this viewpoint, we proposed a kind of new matrix factorization based algorithm, term self-regularized latent K-means (SRLKM), for subspace segmentation. In SRLKM, we devised a new graph-regularizer by using one of the solution to SRLKM itself. For further clarification about our proposed method, we declared the relationships between SRLKM and some close-related algorithms. Moreover, an optimization algorithm for solving SRLKM problem was provided. Finally, extensive experiments were conducted to show that SRLKM was superior to the related algorithms.