Integrating AI in our LOFAR pipelines

Title: Integrating AI in LOFAR pipelines
Speaker: Dr. Jurjen de Jong – Leiden University

 https://www.universiteitleiden.nl/en/staffmembers/jurjen-de-jong#tab-1

We are pleased to welcome at our Machine Learning Club, Jurjen de Jong (Postdoctoral Researcher) for a talk on high-resolution LOFAR imaging and the growing role of machine learning in radio astronomy. His work spans AGN, galaxy clusters, wide-field surveys, VLBI, and advanced LOFAR processing workflows. This session will give an accessible and up-to-date overview of how new calibration techniques and AI-based methods are reshaping low-frequency radio astronomy.

Recent developments in LOFAR have made it possible to produce very deep 2.5° × 2.5° images at sub-arcsecond resolution at 150 MHz. These images open a new window for studying radio-loud AGN, star-forming galaxies, supernova remnants, and diffuse cluster emission at unprecedented levels of detail. Achieving these resolutions, however, requires a complex calibration chain. One of the biggest challenges is correcting image artefacts introduced by the ionosphere, which can distort the apparent positions and shapes of sources.

Jurjen will discuss the calibration framework used to create these deep fields and explain how machine-learning approaches help reduce the need for manual intervention. These methods allow the processing pipeline to detect and correct calibration issues more reliably, saving significant time during large-scale surveys. He will also introduce new work using generative AI models to create synthetic training data. This approach can reduce computational costs, explore calibration edge cases, and potentially improve overall performance when dealing with challenging ionospheric conditions.

The talk is aimed at students and researchers who are interested in LOFAR data processing, AGN science, machine learning, and the future of automated calibration in radio astronomy. Whether you work directly with LOFAR or are curious about AI in large-scale astronomical surveys, this session will offer clear insights into current methods and future directions.

Abstract:

We have recently produced deep, 2.5°×2.5° LOFAR images at sub-arcsecond
resolution and 150 MHz, opening a new window for studying radio-loud
AGN, galaxy clusters, star-forming galaxies, supernova remnants, and
other radio sources at unprecedented detail. Achieving these resolutions
requires multiple calibration steps, particularly to correct for image
artefacts introduced by the ionosphere. To automate this process, we
have begun introducing machine-learning approaches that replace manual
intervention during calibration. We are now also exploring the use of
generative AI models to create synthetic data that can reduce
computational costs and potentially improve calibration performance. In
this journal club session, I will give a brief overview of our
calibration framework in the context of LOFAR data processing and
outline our methods to automate and improve calibration using machine
learning.

Date / Time: 5th December  2025, 11:15–12:00
Location / MS Teams: OCW 0017 / Join on MS Teams

Presentation slides : AI for LOFAR VLBI calibration