Classification of Compact Stars via Machine Learning and Neural Network Models

Recent advances in multimessenger astronomy, particularly through gravitational-wave observations of compact-object mergers, have significantly improved our understanding of dense matter. Nevertheless, the internal composition of compact stars remains uncertain. Depending on the underlying equation of state (EoS), these objects may be neutron stars composed primarily of nucleons, quark stars made of deconfined quark matter, or hybrid stars containing both hadronic and quark phases. More exotic constituents, such as hyperons, meson condensates, or dark matter, have also been proposed. In this work, we investigate whether the internal composition of compact stars can be inferred from observable quantities, including mass, radius, and tidal deformability. To address this problem, we employ machine-learning and deep-learning techniques trained on a larg dataset of EoSs describing both neutron stars and quark stars. From these EoSs, we generate the corresponding mass radius relations spanning a wide range of stellar configurations. The resulting dataset is used to train and evaluate classification models aimed at identifying the nature of compact objects from their macroscopic properties. Our results indicate that suitable combinations of observables can distinguish neutron stars from quark stars with very high accuracy. These findings demonstrate the potential of machine-learning approaches as tools for probing the composition of dense matter. However, further studies incorporating additional scenarios, including hybrid stars and other exotic forms of matter, are required to establish the robustness and general applicability of this methodology.