Date: February 04, 2026, 04:00 – 05:30 p.m. CET.

Speaker: Prof. Dr. Peter Benner (Max-Planck-Institut Magdeburg)

Location:

     Weyertal 86–90, 50931 Cologne
     Mathematical Institute (Google Maps, OpenStreetMap)
     Seminar Room 1 (Room 0.05)

Title: Computational and Stability-informed Learning of Dynamical Systems

Abstract: We study the derivation of mathematical models of dynamical systems from measured or synthetic (computed) time-domain data (trajectories, time series). 
Particular emphasis is given to learning models satisfying certain physical properties like stability or conservation laws. 
We show how such constraints can be explicitly parameterized in the learning problem in strong form. We will demonstrate this for several methods, including operator inference, sparse identification of dynamical systems (SINDy), and deep learning.