Speaker
Topic
Automating the analysis and design of large-scale optimization algorithms
Abstract
The next generation of complex engineered systems will see an unprecedented integration of electromechanical components, communication, and embedded computation. Imminent examples include self-driving vehicles, smart buildings, and UAVs for automated delivery of goods. It is critical that these new technologies be safe and efficient, as their failure would be socially and economically catastrophic. This talk will focus on the challenge of integrating data-driven optimization algorithms into safety-critical control systems. The problem of selecting a suitable algorithm for use in large-scale optimization is currently more of an art than a science; a great deal of expertise is required to know which algorithms to apply and how to properly tune them. Moreover, there are seldom performance or robustness guarantees. Our key observation is that iterative optimization algorithms may be viewed as discrete-time controllers, and the problem of algorithm selection/tuning may be viewed as a robust control problem. This viewpoint allows us to treat both electromechanical and algorithmic components in a unified manner. By solving simple semidefinite programs, we can derive robust bounds on convergence rates for popular algorithms such as the gradient method, proximal methods, fast/accelerated methods, and operator-splitting methods such as ADMM. Finally, our framework can be used to search for algorithms that meet desired performance guarantees, thus establishing a new and principled methodology for algorithm design. As an illustrative example, we synthesize a new family of first-order algorithms that explore the trade-off between performance and robustness to noise.
Speaker's biography
Laurent Lessard was born and raised in Toronto, Canada. He received the B.A.Sc. degree in Engineering Science from the University of Toronto and the M.S. and Ph.D. degrees in Aeronautics and Astronautics from Stanford University. He is currently a postdoctoral scholar in the Berkeley Center for Control and Identification at the University of California, Berkeley. Before that, he was an LCCC postdoc in the Department of Automatic Control at Lund University in Sweden. His research interests include decentralized control, robust control, and large-scale optimization. Dr. Lessard received the O. Hugo Schuck Best Paper Award at the American Control Conference in 2013.
Invited by the Department of Electrical and Computer Engineering