Deep learning in high angular-resolution radio interferometry

Samira Rezaei Badafshani


253 Downloads (Pure)


This thesis has addressed several challenges of the big data era in the field of high angular-resolution radio astronomy using machine learning algorithms. The methodologies presented in this thesis were designed with the aim of minimizing the need for human interactions, while still providing robust results. This thesis has an interdisciplinary approach and uses knowledge in computer science to advance our understanding of the radio sky. The main objectives of this thesis can be categorized into four subjects. First, it provides an analysis to the properties of the detected radio sources with Very Long Baseline Array (VLBA). Then we have provided the details of our developed source detection and characterization pipeline that can localize the source in any observed image from the VLBA. Beside source detection, the implemented pipeline can remove the observational noise, restore the structure of the celestial sources and predict their properties, such as size and brightness. In the fourth chapter, we have designed an algorithm that can find rare types of galaxies, called strong gravitationally lensed systems, among the many observed radio emitting objects observed with the International LOFAR Telescope. We also have provided preliminary results on using machine learning algorithms to predict the lensing parameters such as the Einstein radius, axis ratio and position angle.
Originele taal-2English
KwalificatieDoctor of Philosophy
Toekennende instantie
  • Rijksuniversiteit Groningen
  • McKean, John, Supervisor
  • Biehl, M. , Supervisor
Datum van toekenning27-jun.-2022
Plaats van publicatie[Groningen]
StatusPublished - 2022


Duik in de onderzoeksthema's van 'Deep learning in high angular-resolution radio interferometry'. Samen vormen ze een unieke vingerafdruk.

Citeer dit