Heidelberg 2022 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

T: Fachverband Teilchenphysik

T 72: Cosmic Ray 3

T 72.5: Talk

Wednesday, March 23, 2022, 17:15–17:30, T-H31

Neural networks for cosmic ray simulations — •Pranav Sampathkumar, Antonio Augusto Alves Junior, Tanguy Pierog, and Ralf Ulrich for the CORSIKA 8 collaboration — Institute for Astroparticle Physics (IAP) - KIT

Simulating cosmic ray showers at high energies is very memory and time intensive. Current model-dependent hybrid techniques are constrained by our ability to model from known physics. This contribution discusses novel machine learning techniques in order to bypass explicit simulations, and extract features which can't be modeled easily from first principles. The potential of Generative Adversarial Neural Networks (GANs) in learning and emulating cosmic ray simulations is discussed, along with a presentation of preliminary attempts in using a GAN in generating universal electron-positron distributions associated to showers with varying primaries and energies. The applicability and potential pitfalls in using a neural network based approach for cosmic ray simulations is also discussed. Finally, a CONEX (hybrid simulations using cascade equations) inspired Recurrent Neural network (RNN) model is presented. Preliminary results obtained from training an RNN using a cosmic ray simulation dataset for electromagnetic cascades generated using CORSIKA8 are summarized.