English

Online nonparametric regression with Sobolev kernels

Statistics Theory 2021-07-14 v2 Machine Learning Machine Learning Statistics Theory

Abstract

In this work we investigate the variation of the online kernelized ridge regression algorithm in the setting of dd-dimensional adversarial nonparametric regression. We derive the regret upper bounds on the classes of Sobolev spaces Wpβ(X)W_{p}^{\beta}(\mathcal{X}), p2,β>dpp\geq 2, \beta>\frac{d}{p}. The upper bounds are supported by the minimax regret analysis, which reveals that in the cases β>d2\beta> \frac{d}{2} or p=p=\infty these rates are (essentially) optimal. Finally, we compare the performance of the kernelized ridge regression forecaster to the known non-parametric forecasters in terms of the regret rates and their computational complexity as well as to the excess risk rates in the setting of statistical (i.i.d.) nonparametric regression.

Cite

@article{arxiv.2102.03594,
  title  = {Online nonparametric regression with Sobolev kernels},
  author = {Oleksandr Zadorozhnyi and Pierre Gaillard and Sebastien Gerschinovitz and Alessandro Rudi},
  journal= {arXiv preprint arXiv:2102.03594},
  year   = {2021}
}

Comments

40 pages, 5 figures, 3 tables (version 2)

R2 v1 2026-06-23T22:54:04.342Z