ONPPn: Orthogonal Neighborhood Preserving Projection with Normalization and its applications
作者：
Highlights：
• Most of the DR techniques find a subspace by either applying orthogonality constraint or normalization constraint.
• Both constraints have their own advantages but combining both constraints increases optimization difficulties.
• Proposed method ONPPn gives an iterative solution to satisfy both the constraints simultaneously.
• The article also proposes 2-dimensional extension of ONPPn.
• Experiment supports that ONPPn improves recognition and reconstruction performance compared to NPP and ONPP.
摘要
•Most of the DR techniques find a subspace by either applying orthogonality constraint or normalization constraint.•Both constraints have their own advantages but combining both constraints increases optimization difficulties.•Proposed method ONPPn gives an iterative solution to satisfy both the constraints simultaneously.•The article also proposes 2-dimensional extension of ONPPn.•Experiment supports that ONPPn improves recognition and reconstruction performance compared to NPP and ONPP.
论文关键词：Dimensionality reduction,Orthogonality,Normalization,Neighborhood preserving projection
论文评审过程：Received 26 October 2017, Revised 17 April 2018, Accepted 8 June 2018, Available online 15 June 2018, Version of Record 28 June 2018.
论文官网地址：https://doi.org/10.1016/j.imavis.2018.06.002
