Programm / Abstract:
For many complex systems the essential dynamical behavior can be uncovered by studying significantly less complex surrogate systems. In many cases these surrogate system can be found by scale separation or dimension reduction techniques. These approaches are based on certain fundamental assumptions, e.g., explicit dependence of the governing equations on asymptotic parameters, large numbers allowing for mean field approaches, or global dominant modes. For many real-world systems these fundamental assumptions are not satisfied and most of the standard approaches for deriving surrogate systems fail. The talk will address the questions of whether, for such complex systems, there still is a low-dimensional manifold supporting the effective dynamics and how to perform efficient numerical simulations of it. In particular, we will demonstrate how to identify this manifold by appropriate embedding techniques and how to find rigorous estimates for the accuracy of this approach by means of a mix of techniques based on transfer operator analysis as well as machine learning.
am Montag den 24. April 2017 um 16:15
TU Berlin, Hauptgebäude
Straße des 17. Juni 135
H3005 3. Etage
eingetragen von Annika Preuss(email@example.com, 314 73620)
zurück zum Kalender Mathematics Calendar of the AMS