BEGIN:VCALENDAR
VERSION:2.0
PRODID:Events Booking
BEGIN:VEVENT
UID:69d6994d74364
SUMMARY:A Free Online VKI Seminar - Reliability of data-driven strategies i
 n control and prediction of complex systems
DESCRIPTION:A Free Online VKI Seminar
 Reliability of data-driven strategi
 es in control and prediction of complex systems.
 Our Guest Speaker: Dr. 
 Onofrio Semeraro\, CNRS Research Associate at the Laboratoire Interdiscipl
 inaire des Sciences du Numérique (LISN) - Universite Paris Saclay\, Orsay
  (France)
 Abstract
 The breakthrough of machine-learning based applicat
 ions is revolutionizing scientific computing\; however\, despite numerous 
 successful examples -- mainly based on deep neural network -- the relative
  simplicity of applications of these tools is flawed by numerous issues th
 at are too often overlooked such as generalizability of the models\, lack 
 of guarantees or case dependency. Increasing the size of the dataset of tr
 aining\, as well as relying on deeper and more expressive architectures do
  not necessarily guarantee a solution for these issues\, while leading to 
 longer training and thus to higher computational costs. The potential lack
  of robustness of these predictions is often mitigated by prior knowledge 
 of physical constraints\, when available. In this talk we will focus on tw
 o applications\, by focusing on reliability and margins of guarantee.
 Fi
 rst\, we will consider Long-Short Term Memory neural networks and thorough
 ly investigate the impact of the training set\, its structure and some iss
 ues relating the use of memory gates on the quality of the long-term predi
 ction. Inspired by ergodic theory\, we analyze the amount of data sufficie
 nt for a priori guaranteeing a faithful model of the physical system\, rel
 ying on a proper dataset design. We show how an informed design of the tra
 ining set\, based on invariants of the system and the structure of the und
 erlying attractor\, significantly improves the resulting models\, opening 
 up avenues for research within the context of active learning. We show tha
 t this learning strategy corresponds at deploying a so-called curriculum\,
  a technique often found in robotics.
 In the second part of the seminar\
 , we will focus on flow control\, a subject that has attracted numerous re
 search efforts in the last decades. Recently\, Reinforcement Learning (RL)
 \, which encompasses a large variety of algorithms and strategies\, has ga
 ined traction as an alternative protocol\, thus circumventing the standard
  approaches involving a preliminary model reduction step. Indeed\, RL algo
 rithms do not require any a-priori knowledge of the equations governing th
 e system to be controlled and solely rely on the local measurements of the
  flow\, based on which a policy is learnt from the interaction of the agen
 t with the environment. Despite successes\, the first documented applicati
 ons of RL for control in fluid dynamics often result in highly non-intuiti
 ve control policies\, also when ``cheaper'' optimal solutions are availabl
 e\; we will discuss the impact of data selection on the convergence as wel
 l as an approach relying on optimistic policy iteration\, where the margin
  of guarantees are explicitly found\, with examples of fluid mechanics int
 erest such as the Ginzburg-Landau system and the linearized boundary layer
  flows.
 Biography
 Onofrio Semeraro received his PhD in Mechanical Engi
 neering at KTH-Stockholm (Sweden) in 2013. He served as postdoctoral resea
 rcher at Ecole-Polytechnique\, Palaiseau (France) and Politecnico of Bari 
 (Italy)\, and he is currently a CNRS Research Associate since 2017\, at th
 e Laboratoire Interdisciplinaire des Sciences du Numérique (LISN) - Unive
 rsite Paris Saclay\, Orsay (France). His studies focus mainly on control\,
  data assimilation\, modelling and data-driven techniques\, ranging from s
 ystem identification to deep learning for fluid mechanics. He is currently
  PI of an ANR project dedicated to optimal control and Reinforcement Learn
 ing for control of fluids\, and contributors for projects at the intersect
 ion of machine learning\, dynamical systems and fluid dynamics.
LOCATION:Sint-Genesius-Rode\, 1640 Flemish Brabant\, Belgium
DTSTART:20240209T090000Z
DTEND:20240209T100000Z
DTSTAMP:20220307T101637Z
ORGANIZER;CN=VKI Secretariat:MAILTO:secretariat@vki.ac.be
GEO:50.73507;4.396005
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Sint-Genesius-Rode\, 1640 F
 lemish Brabant\, Belgium;X-APPLE-RADIUS=72;X-TITLE=von Karman Institute fo
 r Fluid Dynamics:50.73507;4.396005
END:VEVENT
END:VCALENDAR