Kernel Based Transfer Function Estimation with Enhanced Prior Knowledge

28 oktober 2021, Leuven & online
AI for Times Series seminarie, by John Lataire (VUB-Elec)
VAIA, Flanders AI Research & KU Leuven STADIUS

This talk explores opportunities in combining insights from systems theory with tools from machine learning, for data driven modelling of dynamic systems.

The systems and control community has a long history of successful applications in engineering domains, traditionally making use of domain-specific knowledge and dynamic properties of systems. Machine learning has specialised in classification and regression from a black-box perspective, often using a Bayesian approach. Combining both offers opportunities, especially in efficiently optimising model structures and complexities for estimating models of measured dynamic systems, while imposing (interpretable) properties to these models.

This talk discusses the data-driven modelling of Linear Time-Invariant (LTI) systems with Gaussian Processes (GP) regression. First, GP regression in general, and Time domain and frequency domain expressions of LTI systems are reviewed. Then, relevant system-specific properties, s.a. causality and stability are encoded in kernels, used as prior knowledge for estimation purposes. Care is given to visual interpretations of this prior knowledge in the spectral domain. Finally, recent results are shared on the estimation of the more challenging situation, where the systems are lightly damped. The combination of the non-parametric local rational model (LRM) estimator with the GP regression approach is proposed.

John Lataire was born in Brussels, Belgium, in 1983. He received the Electrical Engineer degree in electronics and information processing and the Ph.D. degree in engineering sciences (Doctor in de Ingenieurswetenschappen) from the Vrije Universiteit Brussel, Brussels, in 2006 and 2011, respectively. 

From October 2007 to October 2011, he was on a Ph.D. fellowship from the Research Foundation—Flanders (FWO). Since August 2006, he has been working as a researcher with the Department ELEC-VUB, Brussels. 

Dr. Lataire is the coauthor of more than 40 articles in refereed international journals. He received the 2008 IOP outstanding paper award (best paper in Measurement Science & Technology), the Best Junior Presentation Award 2010 at the 29th Benelux Meeting on Systems and Control, was the co-recipient of the 2014 Andy Chi award (best paper in IEEE Trans. on Instrumentation and Measurement), and was the recipient of the 2016 J. Barry Oakes Advancement Award (from the IEEE Instrumentation and Measurement society).

His main interests include the frequency domain formulation of the identification of dynamic systems, with a specific focus on the identification of time-varying systems, and the use of kernel-based regression in system identification.

Praktisch

Inschrijven?

  • Prijs: gratis
  • Inschrijven is niet nodig, je kan gewoon naar de (online) lezing komen.
    Als je graag een reminder ontvangt, geef dan je contactgegevens hieronder op, dan ontvang je een melding op de dag van het seminarie.

AI for Time Series Seminars

Verscheidene onderzoeksgroepen in het Vlaams AI-onderzoeksprogramma verrichten onderzoek van wereldklasse in verband met tijdreeksen, zowel voor de ontwikkeling van algoritmes en tools, als voor een brede reeks toepassingen. In een recente rondvraag bij de Vlaamse AI-gemeenschap, bleek dat het onderwerp ‘tijdreeksen’ het meest gevraagd werd om toekomstige workshops en cursussen over te organiseren. Met deze seminariereeks komen we aan die vraag tegemoet en brengen we onderzoekers die geïnteresseerd zijn in, of onderzoek verrichten naar, tijdreeksen samen. We bieden hen en andere belangstellenden een gevarieerd programma met nationale en internationale sprekers.

Klaar om te starten?

Herinner me aan de AI for Time Series Seminars

Laat hier je contactgegevens achter om reminders aan de AI for Time Series seminaries te ontvangen.
We gebruiken deze gegevens alleen voor dit doel (lees ook onze Privacyverklaring & disclaimer). Je kan jezelf steeds uitschrijven via de uitschrijflink onderaan elke reminder.