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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 3: Neural Networks I

AKPIK 3.3: Vortrag

Mittwoch, 22. März 2023, 14:30–14:45, ZEU/0118

Reconstructing jet characteristics using neural networks — •Arne Poggenpohl and Felix Geyer — Astroparticle Physics, TU Dortmund University, Germany

Active galactic nuclei (AGN) are among the most observed objects in the nocturnal sky. Several of these AGN have the capability to accelerate matter in their nuclei to relativistic velocities, resulting in jets. These are frequently studied sources of radio emission. Analysis of the kinematic characteristics of radio jets can provide information about physical properties of the host galaxy. Previously, this was mostly done by tracking Gaussian components of the jets manually, which is difficult to reproduce. Therefore, the goal of this work is to automatically detect Gaussian components in radio jets using a neural network and thus enable kinematic analysis. Big data sets can thereby be processed, because it is no longer necessary to concentrate on each individual image.

For the necessary object detection, an architecture based on YOLO is used. This architecture consists exclusively of convolutional layers and requires only one pass for the prediction. This allows it to be fast and accurate at the same time.

In this talk, the current state of the work is presented and improvements for the future are pointed out.