AI and the World's Hard Problems · Astronomy & space · arXiv · 2026-10-22
A Hybrid Neural Network Emulator for M1 Neutrino Reaction Rates in Binary Neutron Star Merger Simulations
A hybrid neural network emulator speeds up neutrino reaction-rate calculations for binary neutron star merger simulations by approximating costly two-dimensional reactions while retaining bns_nurates for inexpensive one-dimensional processes. It reproduced opacities within a few percent and accelerated their evaluation by roughly an order of magnitude. In an Athenak merger simulation, it recovered the reference run’s remnant structure, neutrino luminosities, and ejecta composition at approximately half the overall computational cost.