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

T: Fachverband Teilchenphysik

T 21: Experimentelle Methoden der Astroteilchenphysik I

T 21.4: Talk

Monday, March 19, 2018, 16:50–17:05, Z6 - SR 2.013

Substantial improvement in the MAGIC energy reconstruction through machine learning algorithms — •Kazuma Ishio¹, David Paneque¹, Abelardo Moralejo², and Julian Sitarek³ for the MAGIC collaboration — ¹Max-Planck-Institut für Physik — ²Institut de Fisica d'Altes Energies (IFAE), The Barcelona Institute of Science and Technology, Bellaterra (Barcelona), Spain — ³Division of Astrophysics, University of Lodz, Lodz, Poland

The MAGIC telescopes perform gamma-ray astronomy at energies above 50 GeV and extending to about 50 TeV. The energy of the detected gamma ray is estimated with a set of parameters extracted from the shower image on the cameras, and using Look-Up-Tables (LUTs) derived from Monte Carlo simulations. In this talk, I will show that a strategy using random forest (RF) can substantially improve (with respect to LUT) both the energy bias (30% improvement below 100 GeV) and the energy resolution (about 50% improvement above TeV energies). I will show that the choice of the image parameters and the procedure of nesting the RF process across the entire energy range play a crucial role in this improvement.