Orateur
Description
Systematic uncertainties associated to calibration, selection functions and astrophysical effects are dominating the error budget of SNe Ia cosmology. Correction methods applied to account for these systematics, and especially for the complex combination of selection functions and astrophysical variability, are questionable, particularly given the current H0 and Λ tensions for which SNe Ia data are central.
Recently, the ZTF survey has produced a volume-limited sample of more than a thousand SNe Ia, allowing to directly probe the distribution of SNe Ia parameters without being affected by selection effects. However, extending the cosmological analysis to higher redshifts, leveraging the full ZTF DR2 dataset, and combining it with the future LSST data, requires a robust treatment of selection effects.
Using datasets realistically produced using skysurvey, we train a neural network to infer the simulation input parameters. This novel inference method, called SBI, is a promising avenue to solve the complex problem of cosmological inference with SNe Ia data, and thus accurately derive H0, w0 and wa.