Le mardi 26 septembre 2023, Vincent BOUDART présentera l'examen en vue de l’obtention du grade académique de Docteur en Sciences (Collège de doctorat en sciences spatiales) sous la direction de Maxime FAYS et Jean-René CUDELL.

Cette épreuve consistera en la défense publique d’une dissertation intitulée :

“Detection of minute-long Gravitational Wave transients using Deep Learning methods”.

Abstract

On February 11, 2016, the LIGO and Virgo collaborations announced the historic detection of a gravitational wave originating from a binary black hole merger. This achievement relied on precise waveform models and matched filtering techniques. However, longer transient events, encompassing various astrophysical phenomena such as accretion disk instabilities and fallback of matter onto neutron stars, lack accurate models, making matched filtering ineffective. Instead, these events are detected using the template-free excess-power method, which searches for localized power increases in time-frequency space across detectors, often hindered by environmental and instrumental glitches. This thesis proposes the construction of a Convolutional Neural Network (CNN) to identify burst signals and glitches in time-frequency maps. The CNN distinguishes glitch-like pixel clusters, serving as both an anomaly detector and classifier. Extensive analysis reveals that this new algorithm competes with existing burst pipelines while significantly improving processing speed.

Le Jury sera composé de :

Mme Y. NAZE (Présidente), MM. M. BEJGER (Nicolaus Copernicus Astronomical Centre of the Polish Academy of Sciences), J.R. CUDELL (Co-promoteur), M. FAYS (Promoteur), F. ROBINET (IJCLab Orsay), D. SLUSE (Secrétaire).