SMuK 2023 – wissenschaftliches Programm

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

AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz

AKPIK 4: Neural Networks II

AKPIK 4.1: Vortrag

Mittwoch, 22. März 2023, 15:45–16:00, ZEU/0118

Morphological Classification of Radio Galaxies with wGAN-supported Augmentation — •Janis Kummer^1,3, Florian Griese^1,4,5, Lennart Rustige^1,2, Kerstin Borras^2,6, Marcus Brüggen³, Patrick Connor^1,7, Frank Gaede², Gregor Kasieczka⁷, Tobias Knopp^4,5, and Peter Schleper⁷ — ¹Center for Data and Computing in Natural Sciences (CDCS), Hamburg, German — ²Deutsches Elektronen-Synchrotron DESY, Hamburg, German — ³Hamburger Sternwarte, Hamburg, Germany — ⁴University Medical Center Hamburg-Eppendorf, Hamburg, Germany — ⁵Hamburg University of Technology,Hamburg, Germany — ⁶RWTH Aachen University, Aachen, Germany — ⁷Universität Hamburg, Hamburg, German

Supervised deep learning models for the morphological classification of radio galaxies are very important for processing the data of future large radio surveys. However, labelled training data for such models is limited. We demonstrate the use of generative models, specifically a Wasserstein Generative Adversarial Network (wGAN), to generate artificial data for different classes of radio galaxies. Further, we study the impact of augmenting the training data with images from our wGAN on different classification architectures. We find that it is indeed possible to improve models for the morphological classification of radio galaxies with this technique. In addition, fast simulations of radio galaxies with our wGAN are useful to validate new interferometric machine-learning algorithms.