Quantifying Uncertainties in the Neutron Star Equation of State with Gaussian Processes

Understanding the structure of neutron stars requires an accurate nuclear equation of state (EOS), which describes the relationship between the energy density and pressure of matter. Gaussian Process (GP) models provide a flexible, non-parametric approach to modeling the EOS and quantifying its uncertainties. In this talk, we discuss several applications of GPs to model the EOS obtained from many-body perturbation theory calculations with chiral effective field theory interactions as a function of the baryon density and the isospin asymmetry. The trained GP model enables inference of key low-density EOS parameters, including the symmetry energy and its slope at the nuclear saturation density. For beta-equilibrated matter, we discuss results for the proton fraction, pressure, and speed of sound as functions of the baryon density. We also show results for the crust-core transition density as a first step toward the construction of a unified EOS connecting the crust and core. These results are obtained with GPDiff, a novel JAX-based framework that supports automatic differentiation and efficient uncertainty propagation in GP models. It handles correlated uncertainties and enables the systematic evaluation of derivative quantities, such as the pressure and speed of sound. This work highlights GPDiff's potential to connect nuclear theory with neutron star observables through a data-driven EOS with quantified uncertainties.