[EE/CPS Seminar] Apurv Shukla – Differentially Private Online Resource Allocation

EE/CPS Seminar

Speaker: Apurv Shukla
Postdoctoral Associate at Texas A&M University

Title: Differentially Private Online Resource Allocation
Date: Thursday, 21 December 2023
Time: 11am
Venue: EE MMCR (C241)

Abstract:
We consider an online resource allocation problem when the
decision-maker wishes to preserve the privacy of incoming arrivals and
rewards. We present a family of algorithms based on private mean
estimation and primal-dual schemes that achieve optimal utility
guarantees. We establish utility and privacy bounds on the behavior of
algorithms in this family under minimal assumptions on the arrival
process. These results are further corroborated by a set of numerical
experiments that demonstrate the tightness of our bounds.

Biography of the speaker:
Apurv is a Postdoctoral Associate at Texas A&M working with Prof. Le Xie
and Prof. PR Kumar. His research interest lies in learning and control
with applications in power systems. He obtained his PhD from Columbia
University and his bachelor’s degree from IIT Kharagpur.