English

Lower Bounds for Private Estimation of Gaussian Covariance Matrices under All Reasonable Parameter Regimes

Data Structures and Algorithms 2024-04-30 v1 Cryptography and Security Machine Learning Machine Learning

Abstract

We prove lower bounds on the number of samples needed to privately estimate the covariance matrix of a Gaussian distribution. Our bounds match existing upper bounds in the widest known setting of parameters. Our analysis relies on the Stein-Haff identity, an extension of the classical Stein's identity used in previous fingerprinting lemma arguments.

Keywords

Cite

@article{arxiv.2404.17714,
  title  = {Lower Bounds for Private Estimation of Gaussian Covariance Matrices under All Reasonable Parameter Regimes},
  author = {Victor S. Portella and Nick Harvey},
  journal= {arXiv preprint arXiv:2404.17714},
  year   = {2024}
}

Comments

27 pages, preprint