Bayesian Deep Learning for Cosmology with Euclid

This project will focus on the combination of cutting-edge Deep Learning and statistical methods to enable cosmological inference from next-generation facilities such as the European Space Agency's Euclid mission.

The Euclid satellite, launched in July 2023, is poised to change our understanding of the cosmos with its unprecedented amount of data collected from billions of galaxies. Euclid has the potential to pin down the true nature of some of the most elusive components of our cosmological model, such as: Dark Energy and the nature of gravity on cosmological scales, Dark Matter and neutrinos.

The PhD candidate will join the Euclid Consortium and develop novel methodologies for the analysis of the Euclid dataset. The unprecedented size of this dataset makes it ideal for applications of Deep Learning techniques. Specifically, the candidate will work on combining Deep Learning with state-of-the-art statistical inference pipelines, to enable their scaling to the unprecedented size of the Euclid dataset. The candidate will also lead the application of their new methodologies to Euclid data, which will lead to extremely high-impact constraints on our cosmological model.

Topics explored as part of the PhD the research will include: (Geometric) Deep Learning, Bayesian inference and hierarchical modelling, simulation-based inference, differentiable emulation.