Speaker
Description
The current era of precision cosmology has established the ΛCDM model as a remarkably successful framework for describing observations ranging from the Cosmic Microwave Background to the large-scale structure of the Universe. At the same time, persistent tensions among cosmological probes and unresolved questions, including the nature of dark energy, modified gravity scenarios, and the absolute neutrino mass scale, motivate the development of new approaches capable of extracting additional information from cosmological data.
Upcoming surveys such as DESI, Euclid, Rubin, and Roman will dramatically increase both the precision and volume of cosmological observations, shifting the challenge from data acquisition toward accurate theoretical modeling and inference. A significant fraction of cosmological information is encoded in nonlinear structure formation, where effects from massive neutrinos, galaxy bias, baryonic physics, and possible extensions beyond ΛCDM become increasingly important and highly degenerate.
In this talk, I will discuss recent developments in modeling and extracting cosmological information from nonlinear large-scale structure, with emphasis on galaxy bias and its interplay with cosmological parameters such as neutrino masses. I will then present recent work using machine learning and generative approaches, including diffusion models, to model nonlinear structure formation in cosmologies with massive neutrinos and modified gravity. Particular attention will be given to the question of whether these methods are learning genuine cosmological information beyond traditional summary statistics, opening new opportunities for field-level inference in the era of next-generation surveys.