Fully adaptive dictionary for online correntropy kernel learning using proximal methods

作者：

Highlights：

• A fully adaptive dictionary implies better generalization performance.

• Proximal methods promote a higher sparsity level in online kernel learning.

• Correntropy is suitable for the development of outlier-robust adaptive filters.

• The FADOS-CKL is suitable for large scale and non-Gaussian data processing.

摘要

•A fully adaptive dictionary implies better generalization performance.•Proximal methods promote a higher sparsity level in online kernel learning.•Correntropy is suitable for the development of outlier-robust adaptive filters.•The FADOS-CKL is suitable for large scale and non-Gaussian data processing.

论文关键词：Correntropy,Kernel learning,Nonlinear systems,Proximal operator,Recursive estimation,Sparsity

论文评审过程：Received 25 April 2020, Revised 10 February 2021, Accepted 28 March 2021, Available online 2 April 2021, Version of Record 20 April 2021.

论文官网地址：https://doi.org/10.1016/j.eswa.2021.114976