Tuesday 13 July 2021 14:00 – 15:00 (CEST)
Uppsala University, Sweden
Formulating flexible probabilistic models
One of the key lessons to take away from contemporary machine learning is that flexible models offer the best predictive performance. This has implications in many situations, including system identification. In this lecture I will try to make this concrete by looking at a few constructions that we are working with. In particular I will focus on certain combinations of probabilistic models and deep learning for dynamic phenomena of the type we study in system identification. It is interesting to note that while deep learning-based classification is generally addressed using standardized approaches, this is really not the case when it comes to the study of regression problems. There are currently several different approaches used for regression and there is still room for innovation.
Thomas B. Schön is the Beijer Professor Artificial Intelligence in the Department of Information Technology at Uppsala University. He received the PhD degree in Automatic Control in Feb. 2006, the MSc degree in Applied Physics and Electrical Engineering in Sep. 2001, the BSc degree in Business Administration and Economics in Jan. 2001, all from Linköping University. He has held visiting positions with the University of Cambridge (UK), the University of Newcastle (Australia) and Universidad Técnica Federico Santa María (Valparaíso, Chile). In 2018, he was elected to The Royal Swedish Academy of Engineering Sciences (IVA) and The Royal Society of Sciences at Uppsala. He received the Tage Erlander prize for natural sciences and technology in 2017 and the Arnberg prize in 2016, both awarded by the Royal Swedish Academy of Sciences (KVA). He was awarded the Automatica Best Paper Prize in 2014, and in 2013 he received the best PhD thesis award by The European Association for Signal Processing. He received the best teacher award at the Institute of Technology, Linköping University in 2009. He is a Senior member of the IEEE and a fellow of the ELLIS society.
Thursday 15 July 2021 14:00 – 15:00 (CEST)
The Chinese University of Hong Kong
Regularized System Identification: The Prior Knowledge Awakens
In control engineering practice, the users of system identification may have prior knowledge of control plants to be identified. In order to get better model estimates and further improve the modeling efficiency, the users have thus been suggested to make intelligent choice of experiment design, model set, and identification criterion guided by prior knowledge as well as by observed data. However, it is hard to incorporate prior knowledge of control plants in classical system identification paradigm and there are few results on the systematic use of prior knowledge of control plants in system identification until the birth of the so-called regularized system identification in 2010. The regularized system identification proposes to estimate the impulse response model by a regularized least squares method and its major novelty lies in that with the impulse response model, it is possible to impose an underlying model structure by a carefully designed regularization that incorporates/embeds the prior knowledge of control plants. After ten years of development, it has become a viable paradigm and research frontier of system identification. In this talk, this speaker will share his personal view/experience on the development of the regularized system identification and show how some key problems in regularized system identification can be solved by exploring the “control” characteristics of the prior knowledge of the control plant to be identified. This leads to the view that it is important to develop the regularized system identification methods by making use of the “control” characteristics of the prior knowledge of the control plant to be identified.
Tianshi Chen was born on November 17, 1978, in Heilongjiang, China. He received his Ph.D. degree in Automation and Computer-Aided Engineering from The Chinese University of Hong Kong in December 2008. From April 2009 to December 2015, he was working in the Division of Automatic Control, Department of Electrical Engineering, Linköping University, Linköping, Sweden, first as a Postdoc (April 2009–March 2011) and then as an Assistant Professor (April 2011–December 2015). In December 2015, he returned to China and joined the Chinese University of Hong Kong, Shenzhen, as an Associate Professor.
His research interests are in the field of system identification, machine learning, and automatic control. In the past decade, he has been mainly working in the area of kernel-based regularized system identification and has published in this area over 20 journal papers in peer reviewed international journals. He is/was an Associate Editor for Automatica (2017–present), System & Control Letters (2017–2020), and IEEE Control System Society Conference Editorial Board (2016–2019). In May 2015, he received the Youth Talents Award of the Thousand Talents Plan of China.
Wednesday 14 July 2021 14:00 – 15:00 (CEST)
ABB Corporate Research Västerås, Sweden
An industrial perspective on learning models for decision and control
For an industrial control engineer modelling remains one of the most important skills. For example, when configuring and commissioning Model Predictive Control (MPC) at a process industry often 50 % or more of the project time is spent conducting experiments and estimating the process model. Here most models are obtained via linear black-box identification. For other applications first-principles physical modelling still dominates. In recent years a third candidate has emerged, viz. to apply methods based on Artificial Intelligence and Machine Learning.
This talk will discuss the pros and cons of all these modelling techniques from the perspective of an industrial control researcher. Some application examples will be presented, and an attempt will be made to describe where the more recent AI and ML methods are best utilized and perhaps where they should not be used. Regardless of the chosen modelling technique, in order to reduce the engineering effort it is clear that the modelling procedure needs to be further automated. Some examples of how this can be done will also be discussed.
Alf Isaksson received an MSc in Computer Engineering and a PhD in Automatic Control, in 1983 and 1988 respectively, both from Linköping University, Sweden. After graduating he stayed at Linköping University until 1991 as an Assistant Professor. From 1991 to 1992 he spent one year as a Research Associate at The University of Newcastle, Australia. Returning to Sweden in 1992 Isaksson moved to the Royal Institute of Technology (KTH) in Stockholm, where eventually in 1999 he was promoted to full Professor. During this time he also spent 6 months in 1999 at the University of British Columbia, Vancouver, Canada as visiting professor.
In 2001 he made the shift from academic to industrial research and joined ABB Corporate Research in Västerås, Sweden. After a specialist career culminating in an appointment to Corporate Research Fellow in 2009, from 2012 until June 2020 he held multiple positions responsible for funding and coordinating research inside ABB. Most prominently, from January 2014 until March 2019 he was Group Research Area Manager coordinating all Control research globally at ABB Corporate Research. Meanwhile Isaksson still kept a connection to the academic world as Adjunct Professor in Automatic Control at Linköping University 2006-2015. He is now once again Corporate Research Fellow for Automation and Control.
Friday 16 July 2021 14:00 – 15:00 (CEST)
University of Cambridge, UK
University of California, USA
Quantitative epistemology: conceiving a new human-machine partnership
Quantitative epistemology is a new and transformational area of research pioneered by our lab in Cambridge as a strand of machine learning aimed at understanding, supporting, and improving human decision-making. We are developing machine learning models that capture how humans acquire new information, how they pay attention to such information, how their beliefs may be represented, how their internal models may be structured, how these different levels of knowledge are leveraged in the form of actions, and how such knowledge is learned and updated over time. Because our approach is driven by observational data in studying knowledge as well as using machine learning methods for supporting and improving knowledge acquisition and its impact on decision-making, we call this “quantitative epistemology.”
Our methods are aimed at studying human decision-making, identifying potential suboptimalities in beliefs and decision processes (such as cognitive biases, selective attention, imperfect retention of past experience, etc.), and understanding risk attitudes and their implications for learning and decision-making. This would allow us to construct decision support systems that provide humans with information pertinent to their intended actions, their possible alternatives and counterfactual outcomes, as well as other evidence to empower better decision-making.
Mihaela van der Schaar is the John Humphrey Plummer Professor of Machine Learning, Artificial Intelligence and Medicine at the University of Cambridge, a Fellow at The Alan Turing Institute in London, and a Chancellor’s Professor at UCLA.
Mihaela was elected IEEE Fellow in 2009. She has received numerous awards, including the Oon Prize on Preventative Medicine from the University of Cambridge (2018), a National Science Foundation CAREER Award (2004), 3 IBM Faculty Awards, the IBM Exploratory Stream Analytics Innovation Award, the Philips Make a Difference Award and several best paper awards, including the IEEE Darlington Award.
Mihaela’s work has also led to 35 USA patents (many widely cited and adopted in standards) and 45+ contributions to international standards for which she received 3 International ISO (International Organization for Standardization) Awards.
In 2019, she was identified by National Endowment for Science, Technology and the Arts as the most-cited female AI researcher in the UK. She was also elected as a 2019 “Star in Computer Networking and Communications” by N²Women. Her research expertise spans signal and image processing, communication networks, network science, multimedia, game theory, distributed systems, machine learning and AI.
Mihaela’s research focus is on machine learning, AI and operations research for healthcare and medicine.