Courriel : Mikael.Jacquemont– @ –univ-smb.fr

Téléphone : +33(0) 4 50 09 42 09

Télécopie : +33(0) 4 50 09 65 59

Bureau : A224

Adresse 1 : LISTIC – Polytech Annecy-Chambery, BP 80439, 74944 Annecy le Vieux Cedex, France

Adresse 2 : LAPP, 9 chemin de Bellevue, BP 110, ANNECY LE VIEUX, 74941 ANNECY CEDEX 


Groupe : LISTIC : CIT / LAPP : CTA

Thème : AFUTé

Sujet : Use of deep neural networks for the 3D analysis of gamma ray.

Résumé : The thesis is part of the H2020 ASTERICS European project, aiming to provide solutions to the challenges associated to the processing of the large amount of data from future major astronomical and astroparticle experiments. Especially, the data (several tens of petabytes by 2030) acquired by the Cherenkov Telescope Array (CTA) will require complex and rapid reconstruction methods in order to obtain as much information as possible. This thesis offers to explore new methods based on the recent work developed in Deep Learning to reconstruct the parameters (energy and direction) of the $\gamma$ photons detected by the telescopes of the CTA, providing images in two dimensions. In particular, the thesis will first probe simple convolutional architectures, then multi-tasks learning technics and eventually focus on the use of generative-adversarial approaches. The ability of the implemented methodologies to adapt to other situations (in particular industrial situations) will also be studied.

As a part of the Gamma Learn project, carried by LAPP of CNRS, involving the LISTIC of University Savoie Mont Blanc and Orobix, an Italian company, this thesis takes place in a international collaborative context.

Mots clés : Deep Learning, Rayonnement gamma
Publications : HAL

Encadrant : Patrick Lambert / Gilles Maurin (LAPP)

Co-encadrant : Alexandre Benoit / Thomas Vuillaume (LAPP)

Début de la thèse : octobre 2017

Ecole doctorale : SISEO « Sciences et Ingénierie des Systèmes, de l’Environnement et des Organisations »