English

Estimating the minimizer and the minimum value of a regression function under passive design

Statistics Theory 2023-10-10 v2 Machine Learning Statistics Theory

Abstract

We propose a new method for estimating the minimizer x\boldsymbol{x}^* and the minimum value ff^* of a smooth and strongly convex regression function ff from the observations contaminated by random noise. Our estimator zn\boldsymbol{z}_n of the minimizer x\boldsymbol{x}^* is based on a version of the projected gradient descent with the gradient estimated by a regularized local polynomial algorithm. Next, we propose a two-stage procedure for estimation of the minimum value ff^* of regression function ff. At the first stage, we construct an accurate enough estimator of x\boldsymbol{x}^*, which can be, for example, zn\boldsymbol{z}_n. At the second stage, we estimate the function value at the point obtained in the first stage using a rate optimal nonparametric procedure. We derive non-asymptotic upper bounds for the quadratic risk and optimization error of zn\boldsymbol{z}_n, and for the risk of estimating ff^*. We establish minimax lower bounds showing that, under certain choice of parameters, the proposed algorithms achieve the minimax optimal rates of convergence on the class of smooth and strongly convex functions.

Keywords

Cite

@article{arxiv.2211.16457,
  title  = {Estimating the minimizer and the minimum value of a regression function under passive design},
  author = {Arya Akhavan and Davit Gogolashvili and Alexandre B. Tsybakov},
  journal= {arXiv preprint arXiv:2211.16457},
  year   = {2023}
}

Comments

35 pages

R2 v1 2026-06-28T07:17:08.374Z