# Classifying initial conditions of long GRBs modeled with relativistic radiation hydrodynamics

## TL;DR

This paper introduces a neural network-based method to classify initial conditions of long gamma-ray burst models using light curves from relativistic hydrodynamics simulations, improving parameter estimation accuracy.

## Contribution

It presents a novel approach combining neural networks and hydrodynamics simulations to classify initial conditions of GRBs from their light curves.

## Key findings

- Neural networks can classify GRB initial conditions with high accuracy.
- Refinement of parameter space improves classification precision.
- Method effectively distinguishes key parameters from simulated light curves.

## Abstract

We present a method to classify initial conditions of a long gamma ray bursts model sourced by a single relativistic shock. It is based on the use of artificial neural networks (ANNs) that are trained with light curves (LC) generated with radiation relativistic hydrodynamics simulations. The model we use consists in a single shock with a highly relativistic injected beam into a stratified surrounding medium with profile $1/r^2$. In the process we only consider the bremsstrahlung radiation and Thomson scattering process. The initial conditions we use to train the ANN are three: the rest mass density, Lorentz factor and radiation energy density of the beam that produces the relativistic shock, together with the LC generated during the process. The classification selects the location of a box in the 3d parameter space that better fits a given LC, and in order to decrease the uncertainty of the parameters this box is refined and the classification selects a new box of smaller size.