Efficient Second Order Online Learning by Sketching
Abstract
We propose Sketched Online Newton (SON), an online second order learning algorithm that enjoys substantially improved regret guarantees for ill-conditioned data. SON is an enhanced version of the Online Newton Step, which, via sketching techniques enjoys a running time linear in the dimension and sketch size. We further develop sparse forms of the sketching methods (such as Oja's rule), making the computation linear in the sparsity of features. Together, the algorithm eliminates all computational obstacles in previous second order online learning approaches.
Cite
Text
Luo et al. "Efficient Second Order Online Learning by Sketching." Neural Information Processing Systems, 2016.Markdown
[Luo et al. "Efficient Second Order Online Learning by Sketching." Neural Information Processing Systems, 2016.](https://mlanthology.org/neurips/2016/luo2016neurips-efficient/)BibTeX
@inproceedings{luo2016neurips-efficient,
title = {{Efficient Second Order Online Learning by Sketching}},
author = {Luo, Haipeng and Agarwal, Alekh and Cesa-Bianchi, Nicolò and Langford, John},
booktitle = {Neural Information Processing Systems},
year = {2016},
pages = {902-910},
url = {https://mlanthology.org/neurips/2016/luo2016neurips-efficient/}
}