A nonlinear orthogonal non-negative matrix factorization approach to subspace clustering
作者：
Highlights：
• Subspace clustering is solved from nonlinear orthogonal NMF perspective.
• General kernel-based multiplicative orthogonal updates for NMF are derived.
• Explicit orthogonality constraint excludes the usual k-means clustering step.
• The local geometric structure is included via fully connected graph regularization.
• A connection between spectral clustering and kernel orthogonal NMF is established.
摘要
•Subspace clustering is solved from nonlinear orthogonal NMF perspective.•General kernel-based multiplicative orthogonal updates for NMF are derived.•Explicit orthogonality constraint excludes the usual k-means clustering step.•The local geometric structure is included via fully connected graph regularization.•A connection between spectral clustering and kernel orthogonal NMF is established.
论文关键词：Subspace clustering,Non-negative matrix factorization,Orthogonality,Kernels,Graph regularization
论文评审过程：Received 28 March 2017, Revised 22 March 2018, Accepted 27 April 2018, Available online 2 May 2018, Version of Record 15 June 2018.
论文官网地址：https://doi.org/10.1016/j.patcog.2018.04.029
