conftrace_

2012 NeurIPS NeurIPS 2012

Privacy Aware Learning

Abstract

We study statistical risk minimization problems under a version of privacy in which the data is kept confidential even from the learner. In this local privacy framework, we show sharp upper and lower bounds on the convergence rates of statistical estimation procedures. As a consequence, we exhibit a precise tradeoff between the amount of privacy the data preserves and the utility, measured by convergence rate, of any statistical estimator.

🌉 Interdisciplinary Bridge - Artificial Intelligence and Security & Privacy

📈 Trend Setter - Differential Privacy

🧭 Keyword Pioneer - local privacy

🐣 Hot Topic Early Bird - convergence rate

Authors

Martin J. Wainwright, Michael I. Jordan, John C. Duchi

Topics

Machine Learning > Optimization & Theory > Learning Theory Machine Learning > Optimization & Theory > Statistical Learning Machine Learning > Application Areas > Privacy Machine Learning > Learning Types > Supervised Learning Security & Privacy > Differential Privacy Security & Privacy > Privacy Artificial Intelligence > Core AI > Privacy Machine Learning > Learning Types > Differential Privacy

Keywords

differential privacy privacy-preserving learning privacy preserving statistical estimation privacy aware learning statistical risk minimization local privacy convergence rate privacy-utility trade-off

More from NeurIPS 2012

All NeurIPS 2012 papers →

Bayesian nonparametric models for bipartite graphs

Learning Label Trees for Probabilistic Modelling of Implicit Feedback

Factoring nonnegative matrices with linear programs

Truncation-free Online Variational Inference for Bayesian Nonparametric Models

Provable ICA with Unknown Gaussian Noise, with Implications for Gaussian Mixtures and Autoencoders

Learning Image Descriptors with the Boosting-Trick