The Benelux 2019 will host the following renowned invited speakers:

George J. Pappas

Professor at the Department of Electrical and Systems Engineering, University of Pennsylvania, USA.

Title: Robustness Analysis of Neural Networks via Semidefinite Programming

Abstract: Deep neural networks have dramatically impacted machine learning problems in numerous fields.  Despite these major advances, neural networks are not robust and hence not suitable for safety-critical applications.  In this lecture, we will present a novel framework for analyzing the robustness of deep neural networks against norm-bounded nonlinearities.  In particular, we develop a semidefinite programming (SDP) framework for safety verification and robustness analysis of neural networks with general activation functions. Our main idea is to abstract various properties of activation functions (e.g., monotonicity, bounded slope, bounded values, and repetition across layers) with the formalism of quadratic constraints. We then analyze the safety properties of the abstracted network via the S-procedure and semidefinite programming. Compared to other approaches proposed in the literature, our method is less conservative, especially for deep networks, with an order of magnitude reduction in computational complexity. Furthermore, our approach is applicable to any activation functions.  Such bounds are very important in analyzing the safety of control systems regulated by neural networks.

Short biography: George J. Pappas is the UPS Foundation Professor and Chair of the Department of Electrical and Systems Engineering at the University of Pennsylvania. He also holds a secondary appointment in the Departments of Computer and Information Sciences, and Mechanical Engineering and Applied Mechanics. He is member of the GRASP Lab and the PRECISE Center. He has previously served as the Deputy Dean for Research in the School of Engineering and Applied Science. His research focuses on control theory and in particular, hybrid systems, embedded systems, hierarchical and distributed control systems, with applications to unmanned aerial vehicles, distributed robotics, green buildings, and biomolecular networks. He is a Fellow of IEEE, and has received various awards such as the Antonio Ruberti Young Researcher Prize, the George S. Axelby Award, the O. Hugo Schuck Best Paper Award, the National Science Foundation PECASE, and the George H. Heilmeier Faculty Excellence Award.

Anders Rantzer

Professor at the Department of Automatic Control, Lund University, Sweden.

Title: Towards a Scalable Theory of Control
Classical control theory does not scale well for large systems like traffic networks, power networks and chemical reaction networks. To change this situation, new approaches need to be developed, not only for analysis and synthesis of controllers, but also for modelling and verification. In this lecture we will present some general classes of networked control problems for which scalable distributed controllers can be proved to achieve the same performance as the best centralized ones. We will also show how synthesis and implementation of distributed controllers can be carried out in a scalable manner. Applications in energy networks and traffic will be discussed.

Title:  Adaptive Control - What Can We Learn ?
The history of adaptive control dates back to aircraft autopilot development in the 1950s. Computer control and system identification lead to a surge of research activity during the decades to follow. Recently, research activities have started to grow again, for reasons similar to the growth of machine leaning; abundance of data and computing resources creates an ever-growing stream of engineering opportunities for adaptation. This  presentation will discuss some new research directions for theory of adaptive control. The first one is stimulated by the rapid progress in statistical learning theory for tail and concentration bounds, which makes it possible to analyse transient properties of adaptive controllers in a rigorous manner. The second direction is based on deterministic analysis of worst case response to disturbances. By studying min-max games like in H-infinity control, it is possible to compute adaptive feedback controllers with an optimal exploration-exploitation trade-off.

Short biography: Anders Rantzer received a PhD in 1991 from KTH, Stockholm, Sweden. After postdoctoral positions at KTH and at IMA, University of Minnesota, he joined Lund University in 1993 and was appointed professor of Automatic Control in 1999. The academic year of 2004/05 he was visiting associate faculty member at Caltech and 2015/16 he was Taylor Family Distinguished Visiting Professor at University of Minnesota. During 2008-18 he coordinated the Linnaeus center LCCC at Lund University and he currently serves as head of department. He is a Fellow of IEEE, a member of the Royal Swedish Academy of Engineering Sciences and past chairman of the Swedish Scientific Council for Natural and Engineering Sciences. His research interests are in modeling, analysis and synthesis of control systems, with particular attention to uncertainty, optimization, scalability and adaptation.

Minicourse : Moritz Diehl

Professor of Systems, Control and Optimization at the University of Freiburg, Germany.

Title: A Survey of Generalized Gauss Newton and Sequential Convex Programming Methods, Their Convergence Properties, and Their Applications in Learning and Control

Abstract: This overview talk regards a large class of Newton-type algorithms for nonlinear optimization problems that can exploit “convex-over-nonlinear” substructures. All of the considered algorithms are generalizations of the Gauss Newton method, and all of them sequentially solve convex optimization problems that are based on linearizations of the nonlinear problem functions, and nearly all of them show linear local convergence. Because they are popular in different communities, no generally established terminology exists to date. For example,  "the Generalized Gauss-Newton (GGN) method" means different algorithms in computer science and in numerical mathematics.  Another popular name for some algorithms in this class is “Sequential Convex Programming (SCP)”. Aim of this survey talk is an attempt to present and classify all algorithms from this class, investigate and compare their local convergence properties, and to report on their applications in estimation, learning, and control.

Short biography: Moritz Diehl was born in Hamburg, Germany, in 1971. He studied physics and mathematics at Heidelberg and Cambridge University from 1993-1999, and received his Ph.D. degree from Heidelberg University in 2001, at the Interdisciplinary Center for Scientific Computing. From 2006 to 2013, he was a professor with the Department of Electrical Engineering, KU Leuven University Belgium, and served as the Principal Investigator of KU Leuven's Optimization in Engineering Center OPTEC. In 2013 he moved to the University of Freiburg, Germany, where he heads the Systems Control and Optimization Laboratory, in the Department of Microsystems Engineering (IMTEK), and is also affiliated to the Department of Mathematics. His research interests are in optimization and control,  spanning from numerical method  development to applications in different branches of engineering, with a focus on embedded and on renewable energy systems.