Heidelberg 2022 – wissenschaftliches Programm

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 4: Deep Learning

AKPIK 4.1: Vortrag

Donnerstag, 24. März 2022, 16:15–16:30, AKPIK-H13

Using Graph Neural Networks for improving Cosmic-Ray Composition Analysis at IceCube Observatory — •Paras Koundal for the IceCube collaboration — Institute for Astroparticle Physics, KIT Karlsruhe, Germany

Graph Neural Networks (GNNs) is one of the most emerging and promising research topics in the field of deep-learning. Described using nodes and edges, graphs allow us to efficiently represent relational data and learn hidden representations of input data to obtain better model-prediction accuracy. The success of GNNs is mainly attributed to their unique ability to represent complex input data in its most natural representation. GNNs have hence accelerated and extended the pattern learning, inference drawing of standard deep-learning architectures. This has also made it possible for faster and more precise analysis in astroparticle physics, enabling new insights from massive volumes of input data. IceCube Neutrino Observatory, a multi-component detector concealed deep under the South Pole ice provides a suitable test-case to implement such methods.

The talk will discuss the GNN-based methods for improving cosmic-ray composition understanding in the transition region from Galactic to extragalactic sources, at IceCube Observatory. The implementation benefits by using full signal-footprint information, in addition to reconstructed cosmic-ray air shower parameters. The talk will also explain improvement to individual GNN based model by ensemble methods. The implementation will reduce the time and computing cost for performing cosmic-ray composition analysis while boosting sensitivity.