SECRET (Stochasticity Emulator for Cosmic Ray Electrons) models the flux of cosmic-ray electrons using machine-learning density estimation techniques. The code represents the joint probability distribution of electron spectra across cosmic-ray transport parameter space using masked autoregressive models. This approach enables interpolation of predicted spectra for different propagation conditions without running a full numerical simulation for each parameter set. SECRET can therefore accelerate studies of cosmic-ray electron spectra and facilitate comparisons between theoretical models and observational data.

DarkRayNet uses recurrent neural networks (RNNs) to quickly simulate antiprotons, antideuterons, protons and Helium cosmic ray (CR) spectra at Earth for an extensive range of parameters. The corresponding neural networks are trained on GALPROP (ascl:1010.028) simulations. DarkRayNet can also simulate the cosmic ray fluxes for antideuterons; the spectra can be predicted for a signal from dark matter annihilation DM Antideuterons and for secondary emission Secondary Antideuterons.