2DRLPP: Robust two-dimensional locality preserving projection with regularization
作者：
Highlights：
• Robust two-dimensional locality preserving projection (2DRLPP) is proposed for noisy image recognition.
• 2DRLPP extracts projections from matrix space via the L1-norm criterion, which is robust to outliers.
• The orthogonality constraint is imposed on 2DRLPP to ensure its orthogonal projections.
• A regularization term is adopted to control the model complexity and guarantee the stability of solution.
• An efficient iterative algorithm is presented to solve the L1-norm problem of 2DRLPP, whose convergence is guaranteed theoretically.
摘要
•Robust two-dimensional locality preserving projection (2DRLPP) is proposed for noisy image recognition.•2DRLPP extracts projections from matrix space via the L1-norm criterion, which is robust to outliers.•The orthogonality constraint is imposed on 2DRLPP to ensure its orthogonal projections.•A regularization term is adopted to control the model complexity and guarantee the stability of solution.•An efficient iterative algorithm is presented to solve the L1-norm problem of 2DRLPP, whose convergence is guaranteed theoretically.
论文关键词：Two-dimensional locality preserving projection,L1-norm optimization,Robust modeling,Dimensionality reduction
论文评审过程：Received 13 September 2018, Revised 19 December 2018, Accepted 19 January 2019, Available online 1 February 2019, Version of Record 18 February 2019.
论文官网地址：https://doi.org/10.1016/j.knosys.2019.01.022
