Speaker
Lasse Vuursteen
Post-doctoral Researcher
Department of Statistics
University of Pennsylvania
Date
Friday, February 9, 2024
11:00 A.M. – 12:00 P.M. ET
Location
Nguyen Engineering Building, Room 1109
4511 Patriot Circle
Fairfax, Virginia 22030
Differentially Private Hypothesis Testing for High-Dimensional Federated Data
Abstract
In a federated setting, data is not readily available in one "location'', meaning that there is no access to the complete data. This scenario commonly occurs when data is observed or stored at multiple locations, such as hospitals or servers, from which the data cannot be shared in full because of privacy limitations.
This talk is about the theoretical properties of hypothesis testing for federated, high-dimensional data under differential privacy constraints. The theory derived describes optimal performance in such problems by providing impossibility results and methods that attain the theoretically best performance.
Intuitively, testing should offer a lower cost of privacy than estimation of a high-dimensional object like a function. In the talk, I will contrast the theory for testing with recently derived results for estimation, to see to what extent testing is indeed easier.
The talk is based on joint work with T. Tony Cai and Abhinav Chakraborty.
About the Speaker
Lasse Vuursteen is a post-doctoral researcher at the The Wharton School of the University of Pennsylvania, working with T. Tony Cai. His research mainly focuses on distributed methods in high-dimensional and nonparametric statistics, specifically from communication restricted, privacy and computational point of view.
Event Organizer
David Kepplinger
Assistant Professor, Department of Statistics
College of Engineering and Computing
George Mason University
Nicholas Rios
Assistant Professor, Department of Statistics
College of Engineering and Computing
George Mason University