Analysis of the Applicability of Deep Neural Networks on the Generalization of Neutron Star Equations of State

Abstract

The analysis of equations of state models, which describe the matter inside neutron stars, contributes to the understanding of two fundamental pillars of physics, nuclear matter and gravitation. Recent astrophysical observations such as the event titled GW170817, which founded the era of multi-messenger observations, as well as the important measurements established by the Neutron Star Interior Composition Explorer (NICER) of the radius and mass of the compact objects PSR J0030 + 0451 and PSR J0740 + 6620 brought new perspectives on the limitations and inconsistencies between observational data and predictions through the gravity model. Combining the current motivating scenario with the growth of available data and increased computational capacity, the topic has been expanded with the addition of new tools based on machine learning, which have evolved considerably since the mid-2010s. Seeking to contribute to the understanding through a simple and effective representation while maintaining robustness and reliability of its results among the range of complex models existing in the literature, the work under analysis focuses on the application of deep neural networks in the generalization of neutron star state equations, exploring the bases theories of generalized piecewise polytropic formalism, and the construction of a model whose learning method is based on Bayesian probability.