Extending Cosmic Ray Background in Space Experiments using Generative Adversarial Networks

Cosmic rays (CR) reaching telescope detectors in outer space are known
\nto induce glitches and background noise. The presence of CR noise significantly
\ninfluenced Cosmic Microwave Background (CMB) experiments, like Planck and
\nLiteBIRD, which have a long exposition and hard shelling or filtering. In order to
\naddress this challenge, it is imperative to accurately simulate the CR background
\nthroughout the duration of LiteBIRD’s three-year mission. However, state-of-the-art
\nMonte Carlo (MC) simulations are extremely computational expensive, typically
\nrequiring 30 times the simulated period. We present the Cosmic Rays Artificial
\nBackground (CRAB) code, extending MC simulations with Generative Adversarial
\nNetworks (GAN). By leveraging GANs, we can efficiently generate a sufficient
\nnumber of genuine, statistically independent images, unlike traditional noise analysis
\ntechniques combined with template expansion methods.