Joint sample and feature selection via sparse primal and dual LSSVM

作者：

Highlights：

• A joint primal and dual sparse L1-norm PDLSSVM is presented.

• PDLSSVM could be used to do sample and feature selection simultaneously.

• PDLSSVM is solved by ADMM, and the convergence is ensured theoretically.

• Sparse primal LSSVM and sparse dual LSSVM are also demonstrated.

• Experiments show PDLSSVM outperforms other state-of-the-art sparse LSSVM models.

摘要

•A joint primal and dual sparse L1-norm PDLSSVM is presented.•PDLSSVM could be used to do sample and feature selection simultaneously.•PDLSSVM is solved by ADMM, and the convergence is ensured theoretically.•Sparse primal LSSVM and sparse dual LSSVM are also demonstrated.•Experiments show PDLSSVM outperforms other state-of-the-art sparse LSSVM models.

论文关键词：Least squares support vector machine,Feature selection,Sample selection,L1-norm problem,Sparse learning

论文评审过程：Received 4 March 2019, Revised 27 July 2019, Accepted 1 August 2019, Available online 5 August 2019, Version of Record 25 October 2019.

论文官网地址：https://doi.org/10.1016/j.knosys.2019.104915