Dear All,
We are pleased to invite you to the IEOR Seminar by Dr. Ashwin Verma on Tuesday Feb 24, 9:30 am. The details are given below.
TITLE: Optimization and Learning in Networked Systems Under Information and Resource Constraints
SPEAKER: Dr. Ashwin Verma, Purdue University
DATE & TIME: Tuesday, Feb 24, 2026, 9:30 am
VENUE: IEOR 211 (Seminar room), Dept. of IEOR.
ABSTRACT:
In many networked systems, agents must make decisions using information shaped by communication topology, statistical dependence, and resource limitations. In such settings, information structure directly influences algorithm design and achievable guarantees. This talk develops a perspective in which information constraints are treated as an integral part of the optimization and estimation problem.
I begin with state-dependent distributed convex optimization over graphs, where agents adaptively select communication partners based on disagreement, inducing endogenous network evolution. Using a Lyapunov-based contraction analysis that does not rely on fixed graph connectivity assumptions, I establish almost-sure convergence and clarify how adaptive communication policies influence convergence behavior.
I then discuss a reliability-learning problem arising in multi-agent fact checking, where agents observe common claims but have unknown reliability parameters and no access to ground-truth labels. Formulating the estimator as a stochastic approximation scheme, I analyze its limiting mean-field dynamics and establish almost-sure convergence under suitable conditions.
Finally, I briefly outline related results on policy-gradient methods for linear quadratic control under communication power constraints, where transmission budgets are incorporated directly into the learning problem to characterize trade-offs between communication noise and performance.
Across these directions, adaptive communication, statistical dependence, and resource constraints fundamentally shape achievable guarantees in networked optimization and learning.
SPEAKER BIO: Ashwin Verma is a Postdoctoral Researcher in the Department of Electrical and Computer Engineering at Purdue University, working with Prof. Vijay Gupta. His research focuses on optimization and learning in networked systems, with an emphasis on stochastic approximation, distributed algorithms, and decision-making under information and resource constraints. His work includes convergence theory for state-dependent distributed optimization, reliability learning in multi-agent systems, and resource-aware learning formulations for control problems. He received his Ph.D. in Electrical and Computer Engineering from the University of California, San Diego, where he was advised by Prof. Behrouz Touri, and earned his B.Tech. and M.Tech. degrees in Electrical Engineering from IIT Kanpur.