Göttingen 2025 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
GR: Fachverband Gravitation und Relativitätstheorie
GR 4: GW I
GR 4.5: Vortrag
Dienstag, 1. April 2025, 17:35–17:55, ZHG008
Neural Network Assisted Reduced Order Modeling of Black Hole Mergers — •Julian Luca Berg2, Frank Ohme1, and Thomas Wick2 — 1Max Planck Institute for Gravitational Physics, Hannover, Germany — 2Leibniz University Hannover, Germany
Since 2015, the detection of gravitational waves gives us the possibility to study objects in the universe such as black holes and neutron stars. By parameter estimation, we can approximate properties of these objects. This includes the masses, spins, and distances. To perform reliable parameter estimation, it is important to have precise and fast models for the corresponding gravitational waves. One approach to speed up numerical computations is reduced order modeling. In this presentation, an approach by J.S. Hesthaven and S. Ubbiali is applied to gravitational wave models that performs reduced order modeling with neural networks. Therein, a neural network is built that can quickly compute a reduced order model for a given set of parameters such that the solution is still a reliable approximation. Our approach is substantiated with some numerical simulations.
Keywords: neural network; reduced order modeling; parameter estimation; machine learning; black hole mergers