English

Online Distribution Learning with Local Private Constraints

Machine Learning 2024-02-02 v1 Cryptography and Security Data Structures and Algorithms Information Theory math.IT

Abstract

We study the problem of online conditional distribution estimation with \emph{unbounded} label sets under local differential privacy. Let F\mathcal{F} be a distribution-valued function class with unbounded label set. We aim at estimating an \emph{unknown} function fFf\in \mathcal{F} in an online fashion so that at time tt when the context xt\boldsymbol{x}_t is provided we can generate an estimate of f(xt)f(\boldsymbol{x}_t) under KL-divergence knowing only a privatized version of the true labels sampling from f(xt)f(\boldsymbol{x}_t). The ultimate objective is to minimize the cumulative KL-risk of a finite horizon TT. We show that under (ϵ,0)(\epsilon,0)-local differential privacy of the privatized labels, the KL-risk grows as Θ~(1ϵKT)\tilde{\Theta}(\frac{1}{\epsilon}\sqrt{KT}) upto poly-logarithmic factors where K=FK=|\mathcal{F}|. This is in stark contrast to the Θ~(TlogK)\tilde{\Theta}(\sqrt{T\log K}) bound demonstrated by Wu et al. (2023a) for bounded label sets. As a byproduct, our results recover a nearly tight upper bound for the hypothesis selection problem of gopi et al. (2020) established only for the batch setting.

Keywords

Cite

@article{arxiv.2402.00315,
  title  = {Online Distribution Learning with Local Private Constraints},
  author = {Jin Sima and Changlong Wu and Olgica Milenkovic and Wojciech Szpankowski},
  journal= {arXiv preprint arXiv:2402.00315},
  year   = {2024}
}
R2 v1 2026-06-28T14:34:03.545Z