Real-time sufficient dimension reduction through principal least squares support vector machines

作者：

Highlights：

• First approach to real time SVM-based sufficient dimension (SDR).

• Computationally very fast and more accurate SDR than other methods.

• It allows the real-time update by adding data or deleting old data.

摘要

•First approach to real time SVM-based sufficient dimension (SDR).•Computationally very fast and more accurate SDR than other methods.•It allows the real-time update by adding data or deleting old data.

论文关键词：Central subspace,Ladle estimator,Online sliced inverse regression,Principal support vector machines,Streamed data

论文评审过程：Received 6 February 2020, Revised 12 October 2020, Accepted 25 November 2020, Available online 6 December 2020, Version of Record 18 December 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107768