Partenaires
Menu
Rechercher
Navigation
Accueil > Thèses, Stages, Formation et Enseignement > Propositions de thèses 2026 > Machine learning pour l’analyse de données dans les expériences GRAND et HERON
Machine learning pour l’analyse de données dans les expériences GRAND et HERON
par Tristan Beau - 18 novembre
Titre : Machine learning pour l’analyse de données dans les expériences GRAND et HERON
Directrice/directeur de thèse : Olivier Martineau
Co-encadrant.e : Aurélien Benoit-Lévy
Groupe d’accueil : AM3N
Webpage du projet : grand-observatory.org
Collaboration : GRAND
Description :
When an ultra-high-energy cosmic ray enters the Earth’s atmosphere, it produces a cascade of secondary particles (mainly electrons and positrons). These secondary particles, through their relative motion, emit a radio wave that can be detected by antennas such as those of the GRAND experiment—an ambitious project whose prototype, called GRANDProto300, is deployed in the Gobi Desert in China. These complex signals, resulting from the combined emissions of thousands of relativistic particles, can be simulated using Monte Carlo methods such as the ZHaireS simulator. These methods accurately reproduce observations and are therefore an essential tool for GRAND data analysis, though they are very computationally expensive.
The aim of the PhD will be to develop a method to reconstruct the parameters of the incoming cosmic particle (direction, energy, nature, etc.) using a “Simulation-based inference” approach, which will provide robust reconstruction as well as a rigorous estimate of uncertainties. Other applications of this method—such as radio-signal denoising (critical for improving the detection threshold) or cosmic-candidate selection (far rarer than background-noise events)—will also be explored. They will be tested under real conditions on the GRANDProto300 experiment, with possible application to its successor, HERON, which will be deployed in Argentina starting in 2028.
This PhD will be preceded by an internship aimed at improving the machine-learning methods already initiated within the team to rapidly model radio signals at the antenna level.
Lieu(x) de travail : LPNHE / CEA
Déplacements éventuels : Chine Argentine
Stage proposé avant la thèse : Oui
Dans la même rubrique :
- First joint Bayesian field-level inference of density and velocity fields using DESI galaxies and ZTF SN Ia
- Recherche de violation d’invariance de Lorentz et astrophysique des noyaux actifs de galaxies
- Étalonnage des jets, mesures de sections efficaces et extraction d’alpha_S dans ATLAS et au Futur Collisionneur Circulaire au CERN (FCC-ee)
- Analyse des premières données de l’expérience Hyper-Kamiokande - un observatoire unique pour des événements rares dans l’Univers
- From T2K to Hyper-Kamiokande : Precision Measurements with the ND280 Near Detector in the Pursuit of CP Violation
- Indirect dark matter search using gammapy for CTAO and other gamma-ray observatories and commissioning of the NectarCAM cameras
- Higgs physics analysis at the ATLAS experiment in view of the High Luminosity LHC phase and study and optimization of the Vertex Detector for future collider experiments
- Mesure de l’élément V_ub de la matrice CKM et étude de l’universalité de la saveur leptonique dans les transitions du quark b en quark u avec l’expérience LHCb au CERN
- Search for axion-like particles decaying into collimated photon pairs using machine learning with the ATLAS detector at the LHC
- Instrumentation pour la deuxième phase du spectrographe multi-objet DESI
