Heidelberg 2022 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 1: Data Integration & Processing

AKPIK 1.7: Vortrag

Montag, 21. März 2022, 17:45–18:00, AKPIK-H13

Evaluation of deep learning accelerators for the usage in the cosmic ray simulation CORSIKA~8 — •Dominik Baack and Jean-Marco Alameddine for the CORSIKA 8 collaboration — Astroparticle Physics, WG Elsässer, TU Dortmund University, D-44227 Dortmund, Germany

The proliferation of neural networks has led to the acquisition of specialized hardware to accelerate training and application at an increasing number of scientific sites.

To take advantage of this growth, we investigated the extent to which this hardware can be used to accelerate the complex simulation of cosmic particle showers and which parts of the simulation benefit most.

A number of examples based on CORSIKA~8 are presented to illustrate advantages, disadvantages, and limitations in the choice of methods. In particular, the widely used Nvidia accelerators that was used very successfully for ray tracing of optical photons (e.g. Cherenkov light) will be discussed.