Optimal mean two-dimensional principal component analysis with F-norm minimization
作者：
Highlights：
• Our method simultaneously optimizes the projection matrix and mean in the criterion objective.
• Our method directly considers the reconstruction errors of data while most existing L1-norm 2DPCA methods do not.
• Our method is not only robust but also retains 2DPCA’s desirable properties such as rotational invariance.
• We solve the solution by a non-greedy algorithm, which has closed-form solution and good convergence.
摘要
•Our method simultaneously optimizes the projection matrix and mean in the criterion objective.•Our method directly considers the reconstruction errors of data while most existing L1-norm 2DPCA methods do not.•Our method is not only robust but also retains 2DPCA’s desirable properties such as rotational invariance.•We solve the solution by a non-greedy algorithm, which has closed-form solution and good convergence.
论文关键词：Dimensionality reduction,F-norm,2DPCA,Optimized mean
论文评审过程：Received 13 October 2016, Revised 23 February 2017, Accepted 22 March 2017, Available online 29 March 2017, Version of Record 5 April 2017.
论文官网地址：https://doi.org/10.1016/j.patcog.2017.03.026
