Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 2: Parallel Talks
AKPIK 2.5: Vortrag
Montag, 16. März 2026, 17:00–17:15, KS 00.003
Cleaning MOJAVE observations with neural networks — •Christian Arauner1,2, Kevin Schmitz1,2, Anno Knierim1,2, and Tom Gross1,2 — 1TU Dortmund University — 2Lamarr Institute for Machine Learning and Artificial Intelligence
Radio interferometry enables observations of astronomical objects with high angular resolution. However, the inherent incompleteness of the (u,v)-plane sampling leads to significant noise and artifacts in the reconstructed images. Existing state-of-the-art cleaning algorithms effectively mitigate these effects but are computationally demanding, not easily reproducible and not readily scalable to the data volumes anticipated from next-generation radio telescopes. As a promising alternative, neural network-based approaches offer the potential to automate and accelerate the image reconstruction process.
The Python package radionets implements a deep-learning framework for the reconstruction of calibrated data from radio observations. In this approach, ResNets are used to reconstruct the missing values directly in the (u,v)-plane. The validity of the neural network-based approach has been shown with simulated data, now it is possible to apply it to real measurments from the MOJAVE program. The observations of the MOJAVE archive are particular suitable for the development and testing of these neural networks as it comprises a large data set of high-quality data, measured over a long period of time.
In this talk, I will present first observations of the MOJAVE program, reconstructed with radionets.
Keywords: radio astronomy; radio interferometry; neural networks; machine learning; simulation