Speaker
Description
Accurate modeling of cosmic-ray propagation in the Galaxy is a central ingredient in indirect searches for dark matter. However, numerical propagation codes are computationally expensive, limiting their applicability in large parameter scans and inference studies. In this talk, I will present a novel neural network framework that learns the mapping between arbitrary dark matter antiproton or antideuteron injection spectra and the corresponding propagated fluxes at Earth. Our approach is fully agnostic to the underlying injection spectrum and instead directly operates on the spectral shape. The network takes as input both the injection spectrum and the relevant propagation parameters, enabling fast and flexible predictions across a wide parameter space while still accurately reproducing the results of state-of-the-art propagation codes.
