English

Optimal Online Generalized Linear Regression with Stochastic Noise and Its Application to Heteroscedastic Bandits

Machine Learning 2023-03-28 v2 Optimization and Control Machine Learning

Abstract

We study the problem of online generalized linear regression in the stochastic setting, where the label is generated from a generalized linear model with possibly unbounded additive noise. We provide a sharp analysis of the classical follow-the-regularized-leader (FTRL) algorithm to cope with the label noise. More specifically, for σ\sigma-sub-Gaussian label noise, our analysis provides a regret upper bound of O(σ2dlogT)+o(logT)O(\sigma^2 d \log T) + o(\log T), where dd is the dimension of the input vector, TT is the total number of rounds. We also prove a Ω(σ2dlog(T/d))\Omega(\sigma^2d\log(T/d)) lower bound for stochastic online linear regression, which indicates that our upper bound is nearly optimal. In addition, we extend our analysis to a more refined Bernstein noise condition. As an application, we study generalized linear bandits with heteroscedastic noise and propose an algorithm based on FTRL to achieve the first variance-aware regret bound.

Keywords

Cite

@article{arxiv.2202.13603,
  title  = {Optimal Online Generalized Linear Regression with Stochastic Noise and Its Application to Heteroscedastic Bandits},
  author = {Heyang Zhao and Dongruo Zhou and Jiafan He and Quanquan Gu},
  journal= {arXiv preprint arXiv:2202.13603},
  year   = {2023}
}

Comments

27 pages, 3 figures. In this updated version, we have changed the paper title, added new theoretical results on the FTRL algorithm and mainly focused on stochastic online regression. Refer to arXiv:2202.13603v1 for the previous version, which contains more results on heteroscedastic nonlinear bandits