Huanyu Zhang - Ph.D. Thesis Defense

Description

Huanyu Zhang has scheduled his Ph.D. Thesis Defense for Friday April 30, 2021, at 9:00 am. The defense will be held online via Zoom. 

Differentially Private Statistical Inference

Abstract
Making inference from data is at the core of statistics and machine learning. In many settings, data contains sensitive information about individuals. It is critical that our methods protect sensitive information, and meanwhile not preclude our overall goals of statistical analysis.

In this talk, we focus on differentially private (DP) statistical inference. First, we introduce a toolbox of proving lower bounds in DP statistical inference. Our new tools are simple, easy to apply and we use them to establish optimal sample complexity bounds for several tasks. Then we move to the problem of DP hypothesis selection, which can be viewed as a generalization of multi-way hypothesis testing. Our main algorithm is achieved through a reduction to adversarial maximum selection, where we provide a family of algorithms which are near-optimal in the trade-off between the error and the interactivity.

Zhang's Ph.D. Thesis Advisor is Prof. Jayadev Acharya.