Machine Learning-Based Analytical Expressions for Gray-Body Factors and Application to Primordial Black Holes

Abstract

Symbolic Regression (SR) is a machine learning approach that explores the space of mathematical expressions to identify those that best fit a given dataset, balancing both accuracy and simplicity. We apply SR to the study of Gray-Body Factors (GBFs), which play a crucial role in the derivation of Hawking radiation and are recognized for their computational complexity. We explore simple analytical forms for the GBFs of the Schwarzschild Black Hole (BH). We compare the results obtained with different approaches and quantify their consistency with those obtained by solving the Teukolsky equation. As a case study, we apply our pipeline, which we call \texttt{ReGrayssion}, to the study of Primordial Black Holes (PBHs) as Dark Matter (DM) candidates, deriving constraints on the abundance from observations of diffuse extragalactic $\gamma$-ray background. These results highlight the possible role of SR in providing human-interpretable, approximate analytical GBF expressions, offering a new pathway for investigating PBH as a DM candidate.