English

Near-Optimal Bounds for Learning Gaussian Halfspaces with Random Classification Noise

Machine Learning 2023-07-18 v1 Data Structures and Algorithms Statistics Theory Machine Learning Statistics Theory

Abstract

We study the problem of learning general (i.e., not necessarily homogeneous) halfspaces with Random Classification Noise under the Gaussian distribution. We establish nearly-matching algorithmic and Statistical Query (SQ) lower bound results revealing a surprising information-computation gap for this basic problem. Specifically, the sample complexity of this learning problem is Θ~(d/ϵ)\widetilde{\Theta}(d/\epsilon), where dd is the dimension and ϵ\epsilon is the excess error. Our positive result is a computationally efficient learning algorithm with sample complexity O~(d/ϵ+d/(max{p,ϵ})2)\tilde{O}(d/\epsilon + d/(\max\{p, \epsilon\})^2), where pp quantifies the bias of the target halfspace. On the lower bound side, we show that any efficient SQ algorithm (or low-degree test) for the problem requires sample complexity at least Ω(d1/2/(max{p,ϵ})2)\Omega(d^{1/2}/(\max\{p, \epsilon\})^2). Our lower bound suggests that this quadratic dependence on 1/ϵ1/\epsilon is inherent for efficient algorithms.

Keywords

Cite

@article{arxiv.2307.08438,
  title  = {Near-Optimal Bounds for Learning Gaussian Halfspaces with Random Classification Noise},
  author = {Ilias Diakonikolas and Jelena Diakonikolas and Daniel M. Kane and Puqian Wang and Nikos Zarifis},
  journal= {arXiv preprint arXiv:2307.08438},
  year   = {2023}
}
R2 v1 2026-06-28T11:32:23.355Z