Structure, Analysis, and Synthesis of First-Order Algorithms (Dr. Jared Miller, University of Stuttgart)
Systems & Control Seminar
Abstract
Optimization algorithms can be interpreted through the lens of dynamical systems as the interconnection of a stepsize-governing linear system and a set of gradient or subgradient nonlinearities. This dynamical systems formulation allows for the analysis and synthesis of optimization algorithms through the solution of robust control problems. In this work, we use the internal model principle structurally factorize any convergent separable optimization algorithm into a suitable internal model and a core controller. As the key benefit, we reveal that this factorization permits us to synthesize optimization algorithms even if the subgradient information is transmitted over corrupted networks. Computation of these certified-exponentially-convergent networked algorithms is achieved by solving semidefinite programs under bisection. We demonstrate the factorization of existing optimization algorithms and the automated synthesis of new tracking algorithms.
Biographical information
Jared Miller is currently a Postdoctoral Researcher at the Chair of Mathematical Systems Theory at the University of Stuttgart working with Prof. Carsten Scherer. He received his B.S. and M.S. degrees in Electrical Engineering from Northeastern University in 2018, and his Ph.D. degree from Northeastern University in 2023 under the advisorship of Mario Sznaier (Robust Systems Laboratory). He was previously a Postdoctoral Researcher Automatic Control Laboratory (IfA) at ETH Zurich, in the research group of Prof. Roy S. Smith. He is a recipient of the 2020 Chateaubriand Fellowship from the Office for Science Technology of the Embassy of France in the United States. He was given an Outstanding Student Paper award at the IEEE Conference on Decision and Control in 2021 and in 2022, and was a finalist for the Young Author Award at the 2023 IFAC world congress. His research interests include renewable energy integration, convex optimization, safety analysis, and control of nonlinear systems.
Termin
03. März 202616:00 - 17:00
