Program

Following the ERNSI tradition, the workshop will start with a dinner on Sunday 24 September and will end on Wednesday 27 September after lunch. Tuesday afternoon will be reserved for a social program and a banquet will be organized on Tuesday evening.

Program overview

The technical program will include 2 plenary lectures, 1 discussion lecture, 9 regular talks and 3 poster sessions. Please, note that the poster format is limited to A1 (height=84.1cm, width=59.4cm)

For details about the presentations and time slots, please download the ERNSI 2017 Workshop program.

You can also download the poster teaser slides as well as the oral presentation slides hereafter

Bayesian methods in system identification: equivalences, differences, and misunderstanding, Johan Schoukens (Vrije Universiteit Brussel) and Carl Edward Rasmussen (University of Cambridge)

Poster session A

Beyond stochastic gradient descent for large-scale machine learning, Francis Bach (INRIA)

Poster session B

Canonical Correlation Analysis based identification of LPV systems, Roland Tóth (TU Eindhoven)

Multi-armed bandit formulations for identification and control, Cristian R. Rojas (KTH Royal Institute of Technology)

Direct data-driven control of constrained linear systems, Simone Formentin (Politecnico di Milano)

Inverse source estimation problems in magnetostatics, Juliette Leblond (INRIA)

System identification applications in power system stability monitoring, modeling and control, Luigi Vanfretti (Rensselaer Polytechnic Institute)

Poster session C

Distribution-free prediction: perspectives from a recovering bayesian, Dave Zachariah (Uppsala University)

Are right half-plane zeros necessary for inverse response?, Jan Maciejowski (University of Cambridge)

From structurally independent local LTI models to LPV model, Qinghua Zhang (INRIA)

Compactification of reachability row spaces and global likelihood optimization for linear systems, Bernard Hanzon (University College Cork)

Estimating effective connectivity in linear brain network models, Giulia Prando (University of Padova)

Invited plenary talks

The guest speakers for the 2017 ERNSI workshop are

Francis Bach, INRIA Research Director, Head of the Machine Learning Laboratory SIERRA, a joint research team of ENS, INRIA and CNRS

Title: Beyond stochastic gradient descent for large-scale machine learning

Time slot: Monday, September 25, 13h30-14h30

 Abstract: Many machine learning and signal processing problems are traditionally cast as convex optimization problems. A common difficulty in solving these problems is the size of the data, where there are many observations ("large n") and each of these is large ("large p"). In this setting, online algorithms such as stochastic gradient descent which pass over the data only once, are usually preferred over batch algorithms, which require multiple passes over the data.  In this talk, I will show how the smoothness of loss functions may be used to design novel algorithms with improved behavior, both in theory and practice: in the ideal infinite-data setting, an efficient novel Newton-based stochastic approximation algorithm leads to robustness to ill-conditioning, while in the practical finite-data setting, an appropriate combination of batch and online algorithms leads to unexpected behaviors, such as a linear convergence rate for strongly convex problems, with an iteration cost similar to stochastic gradient descent. (joint work with Nicolas Le Roux, Eric Moulines and Mark Schmidt)

Short biography: Francis Bach is a researcher at Inria, leading since 2011 the machine learning project-team, which is part of the Computer Science Department at Ecole Normale Supérieure. He graduated from Ecole Polytechnique in 1997 and completed his Ph.D. in Computer Science at U.C. Berkeley in 2005, working with Professor Michael Jordan. He spent two years in the Mathematical Morphology group at Ecole des Mines de Paris, then he joined the computer vision project-team at Inria/Ecole Normale Supérieure from 2007 to 2010. Francis Bach is primarily interested in machine learning, and especially in graphical models, sparse methods, kernel-based learning, large-scale convex optimization, computer vision and signal processing. He obtained in 2009 a Starting Grant and in 2016 a Consolidator Grant from the European Research Council, and received in 2012 the Inria young researcher prize. In 2015, he was program co-chair of the International Conference in Machine learning (ICML).

Luigi Vanfretti, Associate Professor at the Electrical, Computer & Systems Department of Rensselaer Polytechnic Institute (RPI), Troy, NY, USA

Title: System identification applications in power system stability monitoring, modeling and control

Time slot: Tuesday, September 26, 09h00-10h00

Abstract: Across the world, electrical power systems are undergoing an unprecedented transition from their conventional energy transmission and distribution model where energy flows had well established seasonal power transfer corridors from production (sources) to distribution/consumption (sinks). The transition is largely due to the adoption of Renewable Energy Sources (RES) at different voltage levels and with different power production capabilities whose production variability characteristics can alter energy flows in short time-scales, and for which the electrical grid was not designed nor it is it well equipped to control. Naturally, electrical grid dynamics, even those that existed since the first interconnections and that raise due to energy exchanges between production zones, are become increasingly active and ever-more complex. At the same time, in places where grid dynamics where not of a concern due to grid over-dimensioning, they are now introducing new challenges for the overall operation of a large and widely geographically spread network such as the power systems in North America and Europe. Evidently, the application of system identification techniques offers a unique toolset allowing engineers to characterize important grid properties that can be used to monitor and alarm operators to unwanted dynamics, while at the same time to improve the ability of power system models to capture grid dynamics. This talk gives an overview of over 10 years of work in the speaker’s career of applying system identification methods and tools in electrical power networks across continents. From initial experiences in estimating the overall system modal response using elementary signal processing methods, through the application of least-costly input design methods and the development of models and tools for automated calibration of power system models, to the development of actual control systems to stabilize power networks, the talk aims to give an understanding of one of the most relevant stability characteristics of large interconnected grids, i.e. inter-area oscillations, and how system identification gives a unique toolset for understanding, monitoring and controlling electrical grid dynamics. As electrical grids continue their transition towards complex cyber-physical systems, the system identification community can bring unprecedented value to help different societies to develop cleaner and greener electrical networks, the aim of this talk is not only to show that the development of advanced system identification techniques is important, but the availability of relevant tools and models that enables their application is also of great significance. Examples of these results include real-time monitoring and power system model validation tools that have been released as open source software, that will be briefly presented in the talk, and are available at: https://github.com/ALSETLab

Short biography: Luigi Vanfretti (IEEE S'03–M'10–SM'13) received the Electrical Engineering degree from Universidad de San Carlos de Guatemala, Guatemala City, Guatemala, in 2005. He was also a Visiting Researcher with The University of Glasgow, Glasgow, Scotland, in 2005. He obtained the M.Sc. and Ph.D. degrees in electric power engineering from Rensselaer Polytechnic Institute, Troy, NY, USA, in 2007 and 2009, respectively. For his research and teaching work toward his Ph.D. degree, he was awarded the Charles M. Close Award from Rensselaer Polytechnic Institute. He was with KTH Royal Institute of Technology, Stockholm, Sweden, as Assistant 2010-2013), and Associate Professor (Tenured) and Docent (2013-2017/August); where he established the SmarTS Lab and research group. During this period, he was also with Statnett SF, the Norwegian electric power transmission system operator, as consultant (2011 - 2012), and Special Advisor in Strategy and in R&D (2013 – 2016). He joined Rensselaer Polytechnic Institute, Troy, NY in August 2017, to continue to develop his research agenda which aims to apply system identification through the model-based system engineering approach in cyber-physical power systems. He is currently developing his new laboratory and research team ALSETLab: http://alsetlab.com/. Dr. Vanfretti, served from 2009 to 2016 in the IEEE Power Engineering Society (PES) PSDP Working Group on Power System Dynamic Measurements, in different capacities, including as Chair from 2014-016. In addition, from 2009 to 2014, he served as Vice-Chair of the IEEE PES CAMS Task Force on Open Source Software. He is an advocate and evangelist for free/libre and open-source software, member of the Open Source Modelica Consortium (OSCM) and Associate Member of the Free Software Foundation. His research interests are in synchrophasor technology applications; and cyber-physical power system modeling, simulation, stability and control.

Social Program (Tuesday 26 September)

We will travel by bus to the Old Lyon’s district (the bus will leave at 14:15 from Domaine St Joseph) to enjoy a two-hour guided tour organized by the Tourism Office.  This tour will enable us to discover Lyon’s Renaissance heritage as well as its gastronomy. After this tour and some free time, we will meet again at 17:45 in front of the Cathédrale St Jean (the end point of the guided tour) to travel back to Domaine St Joseph.

Running session

As you can see in the program, the sessions start each day at 9AM. Thus, you have plenty of time for running. There are many spots around the Domaine Saint Joseph. So feel free to join the ERNSI runners.