Private Hypothesis Selection

DOI10.1109/TIT.2021.3049802MaRDI QIDQ5001497zbMATH OpenOpenAlexFDO

Authors Mark Bun, Gautam Kamath, Thomas Steinke, Zhiwei Steven Wu

Publication date 22 July 2021

Published in IEEE Transactions on Information Theory (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1905.13229

zbMATH Keywords

application to private distribution-free PAC learningdifferentially private algorithmhypothesis selectionsample complexity

Mathematics Subject Classification ID

Convergence of probability measures (60B10) Nonparametric estimation (62G05) Nonparametric inference (62G99)

Abstract: We provide a differentially private algorithm for hypothesis selection. Given samples from an unknown probability distribution $P$ and a set of $m$ probability distributions $mathcalH$, the goal is to output, in a $varepsilon$-differentially private manner, a distribution from $mathcalH$ whose total variation distance to $P$ is comparable to that of the best such distribution (which we denote by $alpha$). The sample complexity of our basic algorithm is $Oleft(fraclogmalph{a}^2+fraclogmalphavarepsilonight)$, representing a minimal cost for privacy when compared to the non-private algorithm. We also can handle infinite hypothesis classes $mathcalH$ by relaxing to $(varepsilon,delta)$-differential privacy. We apply our hypothesis selection algorithm to give learning algorithms for a number of natural distribution classes, including Gaussians, product distributions, sums of independent random variables, piecewise polynomials, and mixture classes. Our hypothesis selection procedure allows us to generically convert a cover for a class to a learning algorithm, complementing known learning lower bounds which are in terms of the size of the packing number of the class. As the covering and packing numbers are often closely related, for constant $alpha$, our algorithms achieve the optimal sample complexity for many classes of interest. Finally, we describe an application to private distribution-free PAC learning.

Cited in

(5)

• On robustness and local differential privacy
• Improved quantum data analysis
• On the power of private likelihood-ratio tests for goodness-of-fit in frequency tables
• Differentially private sampling from distributions
• Mixtures of Gaussians are privately learnable with a polynomial number of samples

This page was built for publication: Private Hypothesis Selection

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5001497)