BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:AI Center Seminar - AI Fundamentals series - Prof. Lorenzo Rosasco
DTSTART:20261112T140000
DTEND:20261112T150000
DTSTAMP:20261109T170129Z
UID:241efdb0dffc87153a21c09cebfad0308ce6e285236c7f66dd1770d4
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Lorenzo Rosasco\nThis talk is jointly organized by the E
 PFL AI Center\, the TML Lab\, and the ELLIS Lausanne and Genoa Units\, as 
 part of the AI Fundamentals series.\n\nTitle\nNonparametric learning of dy
 namical systems and control\n\nAbstract\nWe consider the problem of learni
 ng from a single finite trajectory of a dynamical system. More precisely\,
  we study discrete-time autonomous stochastic systems defining time-homoge
 neous ergodic Markov processes. We first focus on estimating the optimal o
 ne-step prediction function by nonlinear least squares and develop a found
 ational analysis yielding high-probability guarantees with respect to the 
 invariant measure of the process. This analysis makes explicit how the non
 -independent and non-identically distributed nature of trajectory data mod
 ifies classical supervised learning analysis. It also readily extends to h
 igher-order processes and finite state spaces. We then discuss how the ana
 lysis can be refined using weaker notions of ergodicity and assumptions re
 lating the geometry of the chain to the prediction task. Finally\, we show
  how the same approach extends to learning Koopman operators\, which can i
 n turn be used to control nonlinear dynamical systems. In particular\, we 
 discuss a model predictive control approach for switching dynamical system
 s.\n\nBio\nLorenzo Rosasco is a professor at the University of Genova and 
 a principal investigator at the Italian Institute of Technology (IIT). He 
 is a founder and coordinator of the Machine Learning Genova Center (MaLGa)
  and the Laboratory for Computational and Statistical Learning. He leads t
 he Sustainable Artificial Intelligence and Learning unit at IIT. His resea
 rch focuses on developing efficient machine-learning methods with rigorous
  theoretical foundations. He received his PhD from the University of Genov
 a in 2006\, after holding visiting positions at MIT\, the Toyota Technolog
 ical Institute at Chicago\, and the Johann Radon Institute. From 2006 to 2
 013\, he was a postdoctoral fellow and research scientist in MIT’s Depar
 tment of Brain and Cognitive Sciences. He is an ELLIS Fellow and co-direct
 or of both the ELLIS program “Theory\, Algorithms and Computations of Mo
 dern Learning Systems” and the ELLIS Genoa Unit. He has received several
  awards\, including an ERC Consolidator Grant.
LOCATION:ELE 117 https://plan.epfl.ch/?room==ELE%20117 https://epfl.zoom.u
 s/j/61633194192?pwd=al4tguzbd0fzXyRjvu7fxOCWtf5aCE.1
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR