BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Learning Metastable Dynamics: Application to Molecular Dynamics
DTSTART:20250415T161500
DTEND:20250415T171500
DTSTAMP:20260411T013252Z
UID:67b723590f418cf1d39fc816f4f1fc554b53394620e336bec580a8b6
CATEGORIES:Conferences - Seminars
DESCRIPTION:Feliks Nüske\nMetastablility is a phenomenon which often inhi
 bits the efficient simulation of dynamical systems\, or the generation of 
 samples from high-dimensional probability measures. In particular\, it is 
 frequently encountered in computer simulations of biological macromolecule
 s using molecular dynamics. It is well-known that metastable transitions a
 nd their time scales are encoded in the dominant spectrum of certain trans
 ition operators\, also called Koopman operators. The study of Koopman oper
 ators\, and their data-driven approximation by algorithms like the Extende
 d Dynamic Mode Decomposition (EDMD)\, have gained significant traction in 
 recent years.\nIn this talk\, I will report on recent progress concerning 
 the data-driven analysis of metastable systems using Koopman operators. Fi
 rst\, I will introduce approximation methods based on reproducing kernel H
 ilbert spaces (RKHS)\, which allow the use of rich approximation spaces\, 
 and explain how the resulting large-scale linear problems can be solved ef
 ficiently using random Fourier features (RFF). Second\, I will explain how
  similar ideas can be applied to learn models for the infinitesimal genera
 tor\, which allows for a more detailed system analysis\, including interpo
 lation across statistical ensembles\, or the definition of reduced (coarse
  grained) models.\n 
LOCATION:CM 1 517 https://plan.epfl.ch/?room==CM%201%20517
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR