Nordita Astrophysics Seminars

Improving Cosmological & Astrophysics Inference with Machine Learning for Next-Generation 21-cm Experiments

by Dr Michele Bianco (EPFL)

Europe/Stockholm
Albano 3: 6228 - Mega (22 seats) (Albano Building 3)

Albano 3: 6228 - Mega (22 seats)

Albano Building 3

22
Description

Next-generation radio interferometry arrays will revolutionize our comprehension of fundamental physics by directly observing the presence of neutral hydrogen via 21-cm emission in the early universe during the formation of the primordial collapse structure and the radiative feedback from the first galaxy and stars. The Square Kilometre Array Observatory (SKAO) is an international effort to build the world's largest radio telescope that will produce terabytes of data. This data must be analyzed to separate the cosmological signal from the undesired foreground and instrumental contaminations. Current methods are struggling in this complex process when applied to actual radio interferometry observation of SKA's precursors (e.g. MWA, MeerKAT), underlying the importance of testing and developing new methods on increasingly realistic mock observations. Therefore, a synergy between cosmological and instrumental simulation is required to develop more advanced mock observations and more physically motivated cosmological simulations for radio astronomy. On the other hand, we should develop more dynamic and robust techniques to face the complexity of actual radio interferometry observations. In preparation for the SKA experiment, I have contributed to developing machine learning techniques for foreground mitigation and more physically motivated and performant cosmological simulations for the Epoch of Reionization and the Post-reionization era. In my presentation, I will exemplify the astrophysical, cosmological, computational, and data analysis challenges we will face with the next generation of radio interferometry experiments and my contribution to this goal.