Unsupervised machine learning correlations in EoS of neutron stars

TL;DR

This paper uses unsupervised machine learning to analyze correlations among various nuclear physics models' features and their impact on neutron star equations of state, aiming to improve understanding of model dependencies.

Contribution

It introduces unsupervised learning tools to explore correlations in diverse EoS models, enhancing insights into how model features influence neutron star properties.

Findings

Identified key correlations between model parameters and EoS outcomes.

Demonstrated the potential of machine learning to analyze complex nuclear models.

Provided a framework for understanding model feature impacts on neutron star characteristics.

Abstract

Neutron stars are compact objects of large interest in the nuclear astrophysics community. The extreme conditions present in such systems impose big challenges to our current microscopic models of nuclear structure. Equation of states (EoS) are frequently derived from sophisticated quantum mechanical models, such as: relativistic, non-relativistic and many mean-field approaches. Every single model, in general, contains many parameters such as the NN interaction strength, particle compositions, etc. These are particular features of each model and can be represented by numbers and categories in a machine learning context. Different choices of features will affect EoS properties leading to different macroscopic properties of the star. In this work we analyze a selection of EoS containing a variety of different physics models. One of our objectives is to develop tools that enable a better understanding of the correlations among the different model features and the outcome produced by them when employed to model neutron stars.