A comparative artificial neural networks for Schwarzschild black hole (SBH) radius

Abstract

It is consensus among researchers that the data for the black holes is complicated and extremely non-linear in nature. Therefore, it remains a challenging task for them to predict the key characteristics of concerned black holes accurately. The present work offers artificial neural networks assistance in the context of a choice of training functions for the prediction of astrophysical phenomena like the event horizon and radius of black holes. To be more specific, we considered the Schwarzschild black hole as the simplest solution of Einstein's field equations. The Schwarzschild radius and masses are chosen in the last and first layers of the neural networks model, respectively. Two various training functions namely Levenberg-Marquardt training algorithm (LMTA) and Scaled Conjugate Gradient training algorithms (SCGTA) are used. We have observed that the LMTA achieved significantly lower error rates, suggesting a better fit and stronger learning capabilities from the solar masses of black holes. Furthermore, the close alignment between the ANN-predicted and actual Schwarzschild black hole radius demonstrates the LMTA model holds the ability to generalize effectively across unseen masses of black holes.