Aachen 2019 – wissenschaftliches Programm

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Machine-learning methods and computing in astroparticle physics

AKPIK 3.1: Vortrag

Mittwoch, 27. März 2019, 16:00–16:10, H06

Exploring Optical Properties of Antarctic Ice with IceCube Using Gradient Descent — •Alexander Harnisch for the IceCube collaboration — TU Dortmund

The IceCube Neutrino Observatory detects Cherenkov photons emitted by charged particles passing through the antarctic ice. The properties of light propagation in and around the detector must be well understood to be able to learn about the interacting particles and their sources. This talk gives an overview of how to use in-situ light sources to derive ice model parameters by performing gradient descent optimization instead of grid searches. To be able to do so, it is necessary to differentiate the photon propagation simulation. Here we present a working method to differentiate the simulation with respect to absorption coefficients. The derived gradient is fed to an optimizer which performs gradient descent optimization to find the best fitting coefficients.