Würzburg 2018 – scientific programme

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 1: Arbeitskreis Physik, IT & KI (AKPIK) I

AKPIK 1.5: Talk

Monday, March 19, 2018, 17:45–18:00, 70 - HS 00.107

Convolutional Neural Networks applied to Images of the MAGIC Telescopes — •Philipp Hoffmann and Alicia Fattorini for the MAGIC collaboration — TU Dortmund

MAGIC is a stereoscopic system of two Imaging Air Cherenkov Telescopes (IACT) located at the Rouque de los Muchachos Oberservatory on the Canary Island of La Palma at 2200m above sea level. They operate at an energy range of 50 GeV to 50 TeV.

Convolutional neural networks (CNNs) are used in image recognition and often yield a more accurate image analysis for low quality images than other methods. Applying CNNs to images produced by IACTs is thus of high interest, especially for the reconstruction of low energy events. I will motivate explore and present the usage of CNNs in the context of energy reconstruction in comparison to the standard MAGIC analysis chain.

For the CNN to be a viable analysis method, its outputs have to reproduce the spectra and other properties of well known sources like the crab pulsar and have a agreement between experimental and monte carlo data.