Strongly universally consistent nonparametric regression and classification with privatised data

DOI10.1214/21-EJS1845MaRDI QIDQ2044383zbMATH OpenFDO

Authors Thomas B. Berrett, László Györfi, Harro Walk

Publication date 9 August 2021

Published in Electronic Journal of Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/2011.00216

zbMATH Keywords

classificationlocal differential privacyregression estimateuniversal consistency

Mathematics Subject Classification ID

Nonparametric regression and quantile regression (62G08) Asymptotic properties of nonparametric inference (62G20) Classification and discrimination; cluster analysis (statistical aspects) (62H30) Privacy of data (68P27)

Abstract: In this paper we revisit the classical problem of nonparametric regression, but impose local differential privacy constraints. Under such constraints, the raw data $(X_1,Y_1),ldots,(X_n,Y_n)$, taking values in $mathbb{R}^{d}imesmathbbR$, cannot be directly observed, and all estimators are functions of the randomised output from a suitable privacy mechanism. The statistician is free to choose the form of the privacy mechanism, and here we add Laplace distributed noise to a discretisation of the location of a feature vector $X_i$ and to the value of its response variable $Y_i$. Based on this randomised data, we design a novel estimator of the regression function, which can be viewed as a privatised version of the well-studied partitioning regression estimator. The main result is that the estimator is strongly universally consistent. Our methods and analysis also give rise to a strongly universally consistent binary classification rule for locally differentially private data.

Recommendations

• Privacy aware learning
• Local differential privacy: elbow effect in optimal density estimation and adaptation over Besov ellipsoids
• On density estimation at a fixed point under local differential privacy
• Geometrizing rates of convergence under local differential privacy constraints
• Differentially private empirical risk minimization

Cites work

• A distribution-free theory of nonparametric regression
• A nearest neighbor estimate of the residual variance
• A statistical framework for differential privacy
• Adaptive model selection using empirical complexities
• Extremal mechanisms for local differential privacy
• Fast learning rates for plug-in classifiers
• Geometrizing rates of convergence under local differential privacy constraints
• scientific article; zbMATH DE number 17215 (Why is no real title available?)
• k-ANONYMITY: A MODEL FOR PROTECTING PRIVACY
• Local differential privacy: elbow effect in optimal density estimation and adaptation over Besov ellipsoids
• Local nearest neighbour classification with applications to semi-supervised learning
• Minimax Optimal Procedures for Locally Private Estimation
• Non-Parametric Estimation II. Statistically Equivalent Blocks and Tolerance Regions--The Continuous Case
• On the Rate of Convergence of Local Averaging Plug-In Classification Rules Under a Margin Condition
• Randomized response: a survey technique for eliminating evasive answer bias
• RAPPOR
• Theory of Cryptography

Cited in

(6)

• Multivariate density estimation from privatised data: universal consistency and minimax rates
• Density estimation under local differential privacy and Hellinger loss
• On robustness and local differential privacy
• On nonparametric regression with incomplete data via inverse-weighting and their convergence in L_p norms
• Minimax rate for multivariate data under componentwise local differential privacy constraints
• On rate optimal private regression under local differential privacy

Describes a project that uses

Uses Software

• RAPPOR

This page was built for publication: Strongly universally consistent nonparametric regression and classification with privatised data

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2044383)