Network tomography seeks to infer statistics of a random vector X—such as link-level delays—from measurements of an observable vector Y = PX (e.g., end-to-end path-level delay measurements), where P is a known wide matrix. In this talk, we will focus on estimating the mean of X in a distributed parameter-server–worker setting, in which each worker processes samples for a single coordinate of Y and may behave adversarially. We will begin by examining existing estimation strategies based on data encoding, robust aggregation, and homogenization, and we will show why they are either impractical for network tomography or yield only approximate solutions. Next, we will introduce our novel l1-minimization-based approach, which overcomes these limitations. In particular, we will prove that our method can exactly solve heterogeneous linear systems even under adversarial behavior and sporadic data availability. We will also discuss this algorithm's convergence rate. Finally, we will present empirical results demonstrating that our approach outperforms existing methods in terms of accuracy. If time permits, we will also discuss extensions to tracking and more general optimization scenarios. This is joint work with my Ph.D. students, Swetha Ganesh and Nibedita Roy, my collaborator from IMT Atlantique, Dr. Alexandre Reiffers-Masson, and his PhD student, Vishal Halder.
Gugan Thoppe has been an Assistant Professor in the Computer Science and Automation (CSA) department at the Indian Institute of Science (IISc) since 2019. He also serves as an Associate Researcher at the Robert Bosch Centre, IIT Madras. Gugan earned his PhD in 2016 from the Tata Institute of Fundamental Research (TIFR) in Mumbai. He also has postdoctoral research experience from the Technion Institute of Technology in Israel and Duke University in the USA. His research focuses on reinforcement learning, federated learning, stochastic approximation, and random topology. Gugan's research is supported by grants from CEFIPRA (Indo-French), Walmart (CSR), DST-SERB, and NPCI. His work has been recognized with the Pratiksha Trust Young Investigator Award, the IISc Award for Excellence in Teaching, and the TIFR Award for the Best Ph.D. Thesis. He is also an IEEE Senior member and a member of the ACM India Eminent Speaker Panel.