Karlsruhe 2024 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
T: Fachverband Teilchenphysik
T 16: Data, AI, Computing 1 (anomaly detection)
T 16.5: Talk
Monday, March 4, 2024, 17:00–17:15, Geb. 30.33: MTI
Nested Machine Learning Models for the Cherenkov Telescope Array — Lukas Beiske1,2 and •Maximilian Linhoff1 for the CTA collaboration — 1Astroparticle Physics, WG Rhode/Elsässer, TU Dortmund University, D-44227 Dortmund, Germany — 2Institute for Theoretical Physics IV, PAT, Ruhr University Bochum, D-44780 Bochum, Germany
The Cherenkov Telescope Array (CTA) will be the next-generation ground-based very-high-energy gamma-ray observatory covering an energy range from 20 GeV up to 300 TeV. It will operate tens of Imaging Atmospheric Cherenkov Telescopes (IACTs) on the Canary Island of La Palma (CTA North) and at the Paranal Observatory in Chile (CTA South) once construction and commissioning are finished.
Machine Learning techniques are currently being used to analyze data from IACTs. The tools are used to reconstruct the three main properties of the primary particle: its particle type, energy, and origin. A common approach is to train models on parameters extracted from the shower images observed by the telescopes which in turn give one prediction per telescope image. For events triggering multiple telescopes, these individual predictions can be averaged to obtain a single primary particle prediction for every shower event. However, it is possible to improve these averaged predictions by training a second set of machine learning models using all information available about the shower as seen by the whole telescope array. This talk will show the performance of such nested models for CTA.
Keywords: Gamma-Ray Astronomy; Machine Learning