Cosmological exploitation of neural networks: constraining Ωm from the two-point correlation function of BOSS

In this thesis work we exploited two alternative ML-based techniques to put constraints on the matter density contrast,Ωm, from the two-point correlation function of galaxies in the Baryon Oscillation Spectroscopic Survey (BOSS), which is part of the Sloan Digital Sky Survey (SDSS) III. The considered galaxy sample consists of almost 1.2 millions massive galaxies over an effective area of 9329 square degrees, in a comoving volume of 18.7 cubic Gigaparsecs spanning the redshift range 0.2< z <0.75. The two implemented ML models perform a Multi-Class Classification and a Regression Analysis, and have been trained with a large set of mock two-point correlation function measurements, obtained from log-normal mock catalogues with the same observational selections of the BOSS sample.The constraints we obtained with the Classification and Regression models are Ωm= 0.30±0.03 and Ωm= 0.307±0.006, respectively. These results are remarkably consistent with the ones from Alam et al. (2017), that have been obtained through a standard Bayesian analysis using the same galaxy catalogue.In particular, the Regression ML method implemented in this thesis work provides Ωm constraints that are competitive with the ones obtained by state-of-the-art standard analyses, providing a new, independent confirmation of the standard ΛCold Dark Matter (ΛCDM) cosmological framework.This thesis work demonstrates that supervised ML techniques can be effectively applied to observed cosmological data such as galaxy catalogues, and not only to images or simulations.