February 12, 2021, Zoom

Understanding the interplay of dynamics, data, and optimization is a central focus of many fields including control theory and machine learning. Contracting dynamical systems and convex optimization problems are closely-related notions of “nice” or “well-behaved” settings, in which the main advantages of the linear setting carry over more or less intact. Contraction and convexity are both also closely linked to the theory of a monotone operators, which has recently found application in new machine learning models known as “equilibrium networks”. Meanwhile, computational methods to quantify “robustness” of dynamical systems can be fruitfully applied as “regularisation” to improve generalisation of models fit to data. In this talk, we will explore these links at an introductory level and also through a selection of recent results in nonlinear control design, observer design, system identification, and machine learning.

Ian R. Manchester received the B.E. (Hons 1) and Ph.D. degrees in electrical engineering from the University of New South Wales, Sydney, NSW, Australia, in 2002 and 2006, respectively. He held postdoctoral research positions with Umeå University, Sweden, and the Massachusetts Institute of Technology (MIT). Since 2012, he has been with The University of Sydney, Australia, where he is currently Professor of Mechatronic Engineering, Director of the Australian Centre for field Robotics, and Co-Director of the Sydney Institute for Robotics and Intelligent Systems. His research interests include algorithms for control, learning, and identification of nonlinear dynamical systems, with applications in robotics and biomedical engineering. He is an Associate Editor for the IEEE Robotics and Automation Letters.