TIME: 9h50 - 10h10
COFFEE and CROISSANT
TIME: 10h10 - 10h15
TIME: 10h15 - 11h10 (45 + 10 mins)
SPEAKER: Matteo Viel (SISSA)
TITLE: Probing the Universe with Hydrogen
ABSTRACT: I will review the role of atomic Hydrogen as a tracer of the structure formation processes from high to low redshift, in the post-reionization era. I will try to address some fundamental questions like: can we probe dark matter nature and neutrino masses? can we constrain the geometry and dynamics of our Universe using this tracer? is there evidence of new physics beyond the standard cosmological model? can we test General Relativity? Moreover, I will address the most important astrophysical nuisances that could impact such measurements.
TIME: 11h10 - 11h20
TIME: 11h20 - 12h15 (45 + 10 mins)
SPEAKER: Julien Carron (University of Geneva)
TITLE: CMB (de-)lensing, towards a key cosmological probe
ABSTRACT: Precise extraction of the gravitational lensing signal from the Cosmic Microwave Background is forecast to play a key role for several fundamental physics targets of cosmological observations in the next decade. This includes best constraints on 1) the primordial gravitational wave background, or 2) the measurement of the neutrino mass scale. In the first part of my talk, I will first go through the latest results and challenges in CMB lensing. I will then discuss novel lensing measurement techniques of higher fidelity based on the CMB polarization which we are developing and how they can meet these two targets.
TIME: 12h15 - 12h30
TIME: 12h30 - 13h30 (35 + 25 mins)
SPEAKER: Francois Lanusse (CEA Saclay)
TITLE: Merging deep learning with physical models for the analysis of modern cosmological surveys
ABSTRACT: As we move towards the next generation of cosmological surveys, our field is facing new and outstanding challenges at all levels of scientific analysis, from pixel-level data reduction to cosmological inference. As powerful as Deep Learning (DL) has proven to be in recent years, in most cases a DL approach alone proves to be insufficient to meet these challenges, and is typically plagued by issues including robustness to covariate shifts, interpretability, and proper uncertainty quantification, impeding their exploitation in scientific analysis. In this talk, I will instead advocate for a unified approach merging the robustness and interpretability of physical models, the proper uncertainty quantification provided by a Bayesian framework, and the inference methodologies and computational frameworks brought about by the Deep Learning revolution. In particular, we will see how deep generative models can be embedded within principled physical Bayesian modeling, for instance, to enable Simulation-Based Inference (bypassing the need for analytic likelihoods), or to solve a number of astronomical ill-posed inverse problems ranging from simple image denoising all the way to inferring the distribution of dark matter in the Universe. I will also highlight the power of automatic differentiation frameworks for physical modeling, with applications ranging from fully differentiable N-body simulations to greatly accelerating cosmological posterior inference with Hamiltonian Monte-Carlo.