Seminal
The magazine and atlas of EU-funded researchA new publication from the web and AI team at Scott Charnley

Atlas / France

Marie Skłodowska-Curie Action · Running

SERENEt Machine Learning for Square Kilometre Array Observatory

Coordinator
Universite Paris-Saclay, France
Runs
10 February 2027 to 9 February 2029
Status
Running
EU funding
€242k
Total cost
—
Partners
1 organisations in 1 countries
Type
Marie Skłodowska-Curie Action
Grant agreement
101285301

What the project is doing

In the project's own words, from its summary on CORDIS.

Understanding how the first stars and galaxies transformed the Universe during the Epoch of Reionisation (EoR) is one of the primary science goals of the Square Kilometre Array Observatory (SKAO). Its low-frequency instrument, SKA-Low, will produce unprecedented 3D maps of neutral hydrogen. Still, these faint cosmological signals are buried under astrophysical foregrounds and instrumental artefacts several orders of magnitude stronger.

Current approaches mainly rely on power spectrum analysis and deterministic deep learning methods, both of which face limitations in separating signal from contamination and in propagating uncertainties to astrophysical inference. This project proposes a new probabilistic machine-learning framework for tomographic reconstruction of the 21-cm signal. First, I will generate realistic simulations of interferometric data, combining cosmological models, foregrounds, and instrument effects.

Next, I will benchmark existing deterministic pipelines and extend them with Normalising Flow–based generative models combined with stochastic differential equations to capture the full probability distribution of the signal. Finally, the method will be validated on dedicated LOFAR-EoR observations, ensuring robustness to real data challenges. The expected outcomes include: (i) the first uncertainty-aware 21-cm tomographic reconstructions, (ii) open-access codes, datasets and trained models, and (iii) methodologies ready for application to SKA-Low Science Verification data.

By advancing statistical tools at the interface of cosmology and AI, the project directly supports SKAO’s objectives, maximises the scientific return of European infrastructures, and contributes broadly to data-intensive science. This project lays the foundations of a machine-learning–driven paradigm that will enable SKA-Low to open an unprecedented observational window onto the cosmic dawn and the Epoch of Reionisation.

Topics

Who is involved

OrganisationCityCountryRole
Universite Paris-SaclayGif-Sur-YvetteFranceCoordinator

Source: CORDIS, European Commission. Data of 5 Nov 2026, reused under the Commission's reuse policy. View on CORDIS · Grant DOI