Control of dynamical systems is a fundamental problem with many applications in engineering and the natural sciences. Yet, classical control techniques and their guarantees often require statistical assumptions that do not always hold in practice. In this talk, we present provable control methods based on online convex optimization for environments that can change arbitrarily over time. We start by describing an efficient algorithm and accompanying theoretical guarantees for stabilization and control of unknown systems. Inspired by recent applications, we develop performance guarantees for neural network-based controllers in online environments. To conclude, we discuss connections of our approach to improving mechanical ventilation control.
Bio: Xinyi Chen is a PhD candidate in the Department of Computer Science at Princeton University advised by Prof. Elad Hazan. She is also affiliated with Google AI. Previously, she obtained her undergraduate degree from Princeton in Mathematics, where she received the Middleton Miller Prize. She is a recipient of the NSF Graduate Research Fellowship and a participant of EECS Rising Stars at UC Berkeley.