• deep learning
• detection
• gravitational waves
• localization
• low-latency
• multimessenger
• Virgo

This thesis presents an innovative method based on deep learning that can detect gravitational-wave signals and localize their sky position rapidly.
The importance of low-latency localization of gravitational-wave signals is exemplified by the case of GW170817, the first-ever observed coalescence of a pair of neutron stars. The localization of the source in the sky triggered an extensive follow-up campaign that led to the discovery of the electromagnetic counterpart, marking a new era for multimessenger astronomy.
At present, the data analysis pipelines of the LIGO and Virgo collaboration take few minutes from the arrival of a gravitational-wave signal to send an alert to the astronomical community, due to the computational cost of the algorithms that are used to detect the signal and determine the sky localization.
Reducing this time interval would enable astronomers to extract the maximum astrophysical information from the observation.
Deep learning methods are particularly suitable to be used for low-latency analyses since the computationally intensive stage happens as a preliminary step during the so-called training phase in which the algorithm learns to extract the most important information from the data. Once the training is completed, real time data analyses can be performed orders of magnitude faster than traditional methods, which motivates the use of these techniques for the problem of detection and localization of gravitational-wave signals.