SMuK 2023 – scientific programme

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Neural Networks I

AKPIK 3.4: Talk

Wednesday, March 22, 2023, 14:45–15:00, ZEU/0118

Deep-Learning based Estimation of the Ultra-High Energy Cosmic Ray Spectrum using the Surface Detector of the Pierre Auger Observatory — Ralph Engel, Markus Roth, Darko Veberic, Steffen Hahn, and •Fiona Ellwanger for the Pierre Auger collaboration — Karlsruhe Institute of Technology (IAP), Karlsruhe, Germany

To probe physics beyond the scales of human-made accelerators with cosmic rays demands an accurate knowledge of their energy. Ground-based experiments indirectly reconstruct the primary particle energy from measurements of the emitted fluorescence light or the time-dependent signal of the shower footprint.

At the Pierre Auger Observatory, the shower footprint is measured by a regular hexagonal grid of water-Cherenkov detectors. Since the shower development is a very intricate process, it non-trivial to find hidden patterns in the spatial and temporal distributions of signals. With large simulation datasets, we are able to train neural networks tackling such a problem.

In this work, we present a neural network that gives an estimate on the energy of the primary particle. The precision of the predictions is studied by evaluating the neural networks on a simulated test data set with particular regard to the mass-dependent bias. Systematic differences between simulations and measured data require special attention to possible biases, which are investigated. Methods to correct for these biases are presented. Furthermore, the energy spectrum from corrected neural network predictions is built and compared to published results.