Conformal prediction for uncertainties in the neutron star equation of state

TL;DR

This paper applies conformal prediction, specifically Conformalized Quantile Regression, to neutron star equation of state data, providing reliable, distribution-free uncertainty bands with coverage guarantees.

Contribution

It introduces a novel application of conformal prediction to neutron star EoS data, ensuring robust uncertainty quantification without distributional assumptions.

Findings

Empirical coverage confirms robustness of the conformal prediction bands.

Method successfully applied to posterior samples from Bayesian inference.

Provides reliable uncertainty bands for neutron star mass-radius relations.

Abstract

We study uncertainties in the equation of state of neutron stars using conformal prediction as a distribution-free and model-agnostic method that provides coverage guarantees. In particular, we apply the Conformalized Quantile Regression (CQR) method to posterior samples calculated from Bayesian inference, creating reliable uncertainty bands without assuming a specific form of the underlying distribution. We first construct CQR bands as a postprocessing step to the posterior samples of neutron star mas-radius relations provided by the NMMA collaboration and to Quantum Monte Carlo calculations of pure neutron matter. In all cases, empirical coverage studies confirm the robustness of the method.