Stephen Green, Machine Learning for Gravitational Wave Science, Dip. di Fisica

Convocazione ai seminari dei tre vincitori  ex aequo della procedura selettiva per professore di II fascia -SSD PHYS-02/A (EX SSD FIS/02).

 

Title:  Machine Learning for Gravitational Wave Science

 

Abstract:

Since 2015, the LIGO-Virgo-KAGRA network has reported over 200 gravitational-wave detections from merging black holes and neutron stars. These observations have informed our understanding of black hole astrophysics, cosmology, nuclear physics, and fundamental gravity. In the coming decade, new ground- and space-based observatories will detect far more---and more diverse---sources. Extracting science from these data requires both detailed theoretical predictions and fast, accurate inference. I will describe how simulation-based inference, powered by neural network density estimators, is reshaping gravitational-wave analysis. Once trained on simulated data, these models perform Bayesian inference in seconds. These methods also open the door to improved accuracy when trained on realistic noise. I will highlight recent advances in population studies, eccentric binary analyses, and binary neutron star inference, as well as theoretical work on strong-field gravity and ringdown modeling. These advances are transforming gravitational-wave science today and laying the foundation for tomorrow’s discoveries.