Robust discriminative projection with dynamic graph regularization for feature extraction and classification

作者：

Highlights：

• A new feature extraction method named RDPDG is proposed.

• RDPDG is found to be more robust to complex outliers than other related methods.

• Dynamic graph regularization is used to explore the local structure of outlier data.

• An iterative algorithm is developed to solve RDPDG formulations efficiently.

摘要

•A new feature extraction method named RDPDG is proposed.•RDPDG is found to be more robust to complex outliers than other related methods.•Dynamic graph regularization is used to explore the local structure of outlier data.•An iterative algorithm is developed to solve RDPDG formulations efficiently.

论文关键词：Dimensionality reduction,Feature extraction,Robust discriminative projection,Graph regularization,Classification

论文评审过程：Received 27 January 2022, Revised 24 March 2022, Accepted 27 July 2022, Available online 2 August 2022, Version of Record 12 August 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109563