Speaker
Description
Bayesian inference, widely used in gravitational-wave parameter estimation, crucially depends on the choice of priors. Yet, for mergers involving neutron stars, priors are often chosen in an agnostic way, leaving valuable information from nuclear physics and independent observations of neutron stars unused. We propose to encode information on neutron star physics into data-driven prior distributions constructed with normalizing flows, referred to as neural priors. These priors use information from population analyses, as well as constraints on the nuclear equation of state. Applying this framework to GW170817, GW190425, and GW230529, we demonstrate its ability to yield more informative and physically consistent constraints on source parameters such as mass ratio, tidal deformability, and spins, compared to agnostic priors. Moreover, neural priors naturally provide a principled Bayesian source classification between binary neutron star and neutron star-black hole hypotheses. Our method paves the way for classifying future ambiguous low-mass mergers and for continuously incorporating advances in our understanding of the equation of state into gravitational-wave data analysis.
