The promise of Simulation-based inference for the dark sector

About

In virtue of their size and complexity, cosmological and astrophysical data are rapidly becoming intractable by traditional statistical techniques. Unravelling the twin mysteries of dark energy and dark matter requires new data analysis methods capable of extracting robust and reliable knowledge from upcoming data streams, including the Euclid satellite, the Nancy Grace Roman space telescope and the Vera Rubin observatory. I will focus on recent advances in simulation-based inference, probabilistic diffusion models and self-supervised learning, which together are enabling fast, scalable inference that achieves state-of-the-art verifiable performance in a variety of settings, including supernova type Ia cosmology, photometric redshift estimation and weak lensing calibration. Methodological issues like domain shift and model misspecification will also be addressed.

Bio: Roberto Trotta is professor of theoretical physics at the International School for Advanced Study in Trieste, Italy, where is the Director of the Interdisciplinary Laboratory, and a visiting professor at Imperial College London, where he was a professor of Astrostatistics. His research focuses on cosmology, machine learning and data science, with applications to early universe cosmology, dark matter direct and indirect detection, supernova type Ia and large scale structure data. He was awarded the 2018 Chair Lemaître of the University of Louvain for his work on astrostatistics and the 2020 Annie Maunder medal of the Royal Astronomical Society for his contributions to public engagement.

Coffee served at 10h30

00:00:00 Slide 1

00:01:00 Slide 2

00:01:38 Slide 3

00:02:04 Slide 4

00:02:29 Slide 5

00:02:54 Slide 6

00:03:25 Slide 7

00:03:50 Slide 8

00:04:09 Slide 9