Dr. Linjun Zhang - Rutgers University

Friday, September 17, 2021

Linjun Zhang
Assisant  Professor
Department of Statistics, Rutgers University

Friday, Sep 17th, 2021

2 pm on Zoom (info below)


"The cost of privacy in generalized linear models:
Algorithms and optimal rate of convergence"

Abstract:   In this talk, we introduce differentially private algorithms for parameter estimation in both low-dimensional and high-dimensional sparse generalized linear models (GLMs) by constructing private versions of projected gradient descent. We show that the proposed algorithms are nearly rate-optimal by characterizing their statistical performance and establishing privacy-constrained minimax lower bounds for GLMs. The lower bounds are obtained via a novel technique based on Stein's Lemma that generalizes the tracing attack technique for privacy-constrained lower bounds. This lower bound argument can be of independent interest as it applies to general parametric models. Simulated and real data experiments are conducted to demonstrate the numerical performance of our algorithms


Join Zoom Meeting