Die DPG-Frühjahrstagung in Bonn musste abgesagt werden! Lesen Sie mehr ...

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

T: Fachverband Teilchenphysik

T 103: Cosmic rays IV

T 103.6: Vortrag

Freitag, 3. April 2020, 12:15–12:30, L-3.002

Search for Cosmic Ray Sources Using Graph Convolutional Neural Networks — Teresa Bister, Martin Erdmann, Jonas Glombitza, •Niklas Langner, Josina Schulte, and Marcus Wirtz — III. Physikalisches Institut A, RWTH Aachen

Convolutional neural networks (CNNs) are a promising tool in the search for nearby sources of ultra-high energy cosmic rays (UHECRs). Here, CNNs are used to identify patterns in the arrival directions and energies of UHECRs created by their deflection in the Galactic magnetic field. We investigate graph CNNs, which operate on a graph constructed from the UHECR-features, efficiently utilizing both arrival directions and energies of the UHECRs. First, simple toy simulations of single multiplets consisting of a signal pattern on top of an isotropic background are analyzed as a function of the number of injected UHECRs from a source. Second, astrophysical simulations of many sources taking into account the attenuation in photon fields during the propagation in the extragalactic universe are used to evaluate the sensitivity of the graph networks for varying source densities.