Sparse and kernel OPLS feature extraction based on eigenvalue problem solving

作者：

Highlights：

• We establish the connection between OPLS and reduced-rank regression problem (EVD formulation).

• A novel sparse OPLS is proposed, which enhances the solution obtained following the Procrustres approach.

• We propose a novel sparse Kernel OPLS feature extractor for improved performance, interpretability and efficiency.

摘要

Highlights•We establish the connection between OPLS and reduced-rank regression problem (EVD formulation).•A novel sparse OPLS is proposed, which enhances the solution obtained following the Procrustres approach.•We propose a novel sparse Kernel OPLS feature extractor for improved performance, interpretability and efficiency.

论文关键词：Partial least squares,Orthonormalized PLS,Lasso regularization,Feature extraction,Sparse kernel representation

论文评审过程：Received 30 April 2014, Revised 26 September 2014, Accepted 5 December 2014, Available online 12 December 2014.

论文官网地址：https://doi.org/10.1016/j.patcog.2014.12.002