Detecting Hyperons in neutron stars -- a machine learning approach

TL;DR

This paper introduces a neural network model that predicts the presence of hyperons in neutron stars using observational data like radii and tidal deformabilities, enhancing understanding of neutron star composition.

Contribution

It develops a machine learning approach trained on microscopic equations of state to detect hyperons, tested on real and simulated data for improved neutron star analysis.

Findings

Model accurately predicts hyperon presence with various observational uncertainties.

Neural network performs well on both real and simulated data.

The approach advances neutron star composition detection methods.

Abstract

We present a neural network classification model for detecting the presence of hyperonic degrees of freedom in neutron stars. The models take radii and/or tidal deformabilities as input and give the probability for the presence of hyperons in the neutron star composition. Different numbers of observations and different levels of uncertainty in the neutron star properties are tested. The models have been trained on a dataset of well-calibrated microscopic equations of state of neutron star matter based on a relativistic mean-field formalism. Real data and data generated from a different description of hyperonic matter are used to test the performance of the models.