LANS Seminar

Description: Distributed optimization problems are problems where a group of agents is interested in optimizing a cost function that is decomposable as a sum of local cost functions while each agent is aware of its own local cost function. Many classical approaches for solving these problems require unlimited bandwidth and often ignore the communication constraints between the agents. In the first part of the talk, we study the problem of distributed optimization with imperfect information sharing. We propose and study a two-time-scale decentralized gradient descent algorithm to reach a consensus on an optimizer of the objective function. One time scale fades out the imperfect incoming information from the neighboring agents, and the second one adjusts the local loss functions’ gradients. We show that under certain conditions on the connectivity of the underlying time-varying network and the time-scale sequences, the algorithm converges to an optimal point of the cost function. In the second part of the talk, and on a related topic, we consider distributed optimization over state-dependent networks. Unlike earlier results on this topic, where strong enough assumptions are assumed to ensure convergence, we show that such conditions are unnecessary. We propose and show the convergence of a broad class of algorithms, which we refer to as maximal dissent algorithms. Among many implications, our framework allows for extending the use of load-balancing to distributed optimization with provable convergence guarantees.

Bio: Behrouz Touri is an Associate Professor of the Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana-Champaign. Prior to joining UIUC, he was an Associate Professor of ECE at the University of California, San Diego whre he was a faculty from 2017 to 2025. He received his  B.Sc. degree in Electrical Engineering from Isfahan University of Technology, Isfahan, Iran in 2006, his M.Sc. degree in Communications, Systems, Electronics from Jacobs University, Bremen, Germany in 2008, and his Ph.D. degree in Industrial Engineering from  University of Illinois at Urbana-Champaign in 2011. His research interests include applied probability theory, distributed optimization, control and estimation, and game theory. He was the recipient of the American Control Council’s Donald P. Eckman Award in 2018.

Please note that the meeting URL for this event can be seen on the cels-seminars website which requires an Argonne login.

See all upcoming talks at https://www.anl.gov/mcs/lans-seminars