GECO Cosmo Circle: Optimizing cosmological parameters sampling for deep learning applications of cosmic shear

This Friday the 12th of July at 15:00 we will have the pleasure to hear at Leo Buron (M1 Magistère Physique Fondamentale Université Paris-Saclay) presenting the work carried on so far on his internship supervised by Nicolas Martinet.

Title : “Optimizing cosmological parameters sampling for deep learning applications of cosmic shear”

Abstract: “Weak lensing cosmic shear is a powerful probe to constrain our cosmological model. Modern cosmic shear summary statistics based on mass maps (e.g. higher-order statistics or deep learning) perform even better than traditional two-point statistics but with a cosmological inference based on N-body simulations. These simulations are computationally costly and the sparsity of the parameter space sampling can bias the inference. The aim of this internship is to optimize cosmological parameter sampling, to produce the best predictions for the smallest number of simulations, decreasing both computational time and inaccuracy of the modelling. We test this from fast analytical log-normal simulations with deep learning summary statistics.”

The circle will take place in the Mistral meeting room on the second floor at LAM.