Fast Fisher discriminant analysis with randomized algorithms

作者：

Highlights：

• We propose to use random projection to accelerate Fisher discriminant analysis and provide a theoretical analysis. Empirical study shows our method is effective and efficient.

• We propose to use random feature map to accelerate kernel discriminant analysis. And we provide theoretical and empirical analysis to show the effectiveness and efficiency of our methods.

摘要

•We propose to use random projection to accelerate Fisher discriminant analysis and provide a theoretical analysis. Empirical study shows our method is effective and efficient.•We propose to use random feature map to accelerate kernel discriminant analysis. And we provide theoretical and empirical analysis to show the effectiveness and efficiency of our methods.

论文关键词：Fisher discriminant analysis,Random projection,Random feature map

论文评审过程：Received 31 October 2016, Revised 19 March 2017, Accepted 25 June 2017, Available online 27 June 2017, Version of Record 7 July 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.06.029