Online Algorithms and Applications at the Intersection of Optimization, Control, and Learning

Abstract: Optimization underpins many of the challenges that science and engineering face on a daily basis. Recent years have witnessed a major shift from traditional optimization paradigms grounded on iter-ative algorithms for medium-scale learning and control problems to challenging dynamic, stochastic, and even large-scale settings. This is driven by technological transformations that converted infrastructural, engineering, and social platforms into complex and dynamic systems with human-in-the-loop components and even pervasive sensing and computing capabilities. In this context, this talk presents recent ad-vances in the synthesis and analysis of algorithms for data-driven control and learning with data streams. A time-varying optimization model – that is, problems with costs, constraints, and inputs that evolve over time – is utilized as a common modeling framework to formalize control and learning tasks in dynamic environments and with data streams. The talk then addresses the problem of developing online algorithms that produce decisions in real-time to track optimal solution trajectories of the time-varying problem. When applied to information-processing tasks, the proposed algorithms resolve computational and communication bottlenecks and allow for data processing on the fly. From a control perspective, the proposed approach unlocks opportunities for the development of online algorithms that embed learning applications and are implemented in closed-loop with physical systems to effectively act as feedback controllers and drive the system to optimal operational points. Algorithms are accompanied by an analytical performance analysis in terms of convergence, stability, and tracking. Theoretical and algorithmic foundations can be translated into applications for power systems, transportation networks, and machine learning, and propagate benefits to learning of human-technology interactions and healthcare.

Bio: Emiliano Dall’Anese received the Ph.D. in Information Engineering from the Department of Information Engineering, University of Padova, Italy, in 2011. From January 2009 to September 2010, he was a visiting scholar at the Department of Electrical and Computer Engineering, University of Minnesota, USA, where he was also a Postdoctoral Associate from January 2011 to November 2014. From December 2014 to July 2018 he was a Senior Scientist at the National Renewable Energy Laboratory, Golden, CO, USA. Since August 2018 he has been an Assistant Professor within the Department of Electrical, Computer and Energy Engineering at the University of Colorado Boulder, where he is also an affiliate faculty of the Department of Applied Mathematics. His background spans the broad areas of Optimization, Control, and Signal Processing, and his current research efforts focus on the synthesis of optimization and learning algorithms for decision and information systems. He has received the National Science Foundation CAREER Award in 2020.