Statistics Colloquium Series

Date: 

Monday, September 19, 2022, 12:00pm to 1:00pm

Location: 

Science Center, Room 316

Our upcoming event for the Statistics Department Colloquium Series is scheduled for this Monday, September 19th from 12:00 – 1:00pm (ET) and will be an in-person presentation Science Center Rm. 316. The speaker will be Linjun Zhang who is an Assistant Professor in the Department of Statistics at Rutgers Univeristy.

Title: The Cost of Privacy in Statistical Estimation

Abstract: Privacy-preserving techniques have been a central focus in data analysis nowadays. In this talk, we propose a general technique, the score attack, for lower bounding the differential-privacy-constrained minimax risk of parameter estimation. Inspired by the tracing attack idea in differential privacy, the score attack method is applicable to any statistical model with a well-defined score statistic and capable of optimally lower bounding the minimax risk of estimating unknown model parameters when the estimator is required to be differentially private. The effectiveness of this general method is demonstrated in a variety of examples: the generalized linear model in classical and high-dimensional sparse settings, the Bradley-Terry-Luce model for pairwise comparisons, and non-parametric function estimation in the Sobolev class.