Fast Training Data Generation for Machine Learning Analysis of Cosmic Ray Showers

Applying Machine Learning (ML) methods for the analysis of muon lateral distributions in Extensive Air Showers detected by citizen science projects, while taking into account spatial distribution detectors requires enormous training data sets. Therefore, generating these sets with typical Monte Carlo (MC) generators like CORSIKA is computationally prohibitive. Here we present a method which application special augmentation procedures produces dataset that compatible all essential aspects to produced regular MC computations avoiding their time overhead. We utilize Nakamura-Kamata-Greisen (NKG) was proven be an attractive alternative full-fledged simulations. The simulation $10^{4}$ muons at ground level takes just few seconds using our implementation NKG approach. For notation="LaTeX">$10^{6}$ this figure still around 1 minute. comparison, based performed on Prometheus supercomputer CYFRONET computing center ensemble notation="LaTeX">$\sim 100$ showers initiated particle notation="LaTeX">$10^{16} eV$ resulted 10^{4}$ and 10^{5}$ electrons required computation order days.