Dortmund 2021 – scientific programme

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 2: AKPIK II: Deep Learning

AKPIK 2.3: Talk

Wednesday, March 17, 2021, 16:30–16:45, AKPIKa

Deep Learning based Likelihood Reconstruction of IACT Events — •Noah Biederbeck — TU Dortmund

The Cherenkov Telescope Array (CTA) is the next-generation gamma-ray observatory, currently under construction. Once finished, it will comprise over 100 imaging air Cherenkov telescopes (IACTs) at two sites with the goal of improving over the sensitivity of the current generation by at least an order of magnitude. First prototypes for all telescope variants have achieved first light and are observing.

In this talk a novel approach to event reconstruction for IACT data, based on deep learning, is presented. A generative neural net predicts full camera images from a set of physical event parameters. These predicted images are compared to data using a Poissonian likelihood loss in order to reconstruct the event properties.

First results on single-telescope events will be presented and possible extensions to joint predictions of array events will be outlined.