Dresden 2020 – scientific programme

The DPG Spring Meeting in Dresden had to be cancelled! Read more ...

Parts | Days | Selection | Search | Updates | Downloads | Help

CPP: Fachverband Chemische Physik und Polymerphysik

CPP 42: Data analytics for dynamical systems I (joint session SOE/BP/CPP/DY)

CPP 42.5: Talk

Tuesday, March 17, 2020, 11:00–11:15, GÖR 226

Hyper-Parameter Optimization for Identification of Dynamical Systems — •Tobias Wand¹, Alina Steinberg¹, Tim Kroll², and Oliver Kamps² — ¹Institut für Theoretische Physik, Universität Münster, Deutschland — ²Center for Nonlinear Science, Universität Münster, Deutschland

In recent years, methods to identify dynamical systems from experimental or numerical data have been developed [1,2]. In this context, the construction of sparse models of dynamical systems has been in the focus of interest and has been applied to different problems. These data analysis methods work with hyper-parameters that have to be adjusted to improve the results of the identification procedure. If more than one hyper-parameter has to be fine-tuned, simple methods like grid search are computationally expensive and due to this, sometimes not feasible. In this talk, we will introduce different approaches to optimally select the hyper-parameters for the identification of sparse dynamical systems.
[1] Brunton et al. Proceedings of the National Academy of Sciences, 2016, 113, 3932-3937
[2] Mangan et al. Proceedings of the Royal Society A, 2017, 473, 20170009