Component Separation method for CMB using Convolutional Neural Networks

Abstract: The aim of this project is to recover the CMB anisotropies maps in temperature and polarized intensity by means of a deep convolutional neural network (CNN) which, after appropiate training, can remove the foregrounds from Planck and QUIJOTE data. The results are then compared with those obtained by COMMANDER, based on Bayesian parametric component separation. The CNN successfully recovered the CMB signal for both All Sky and Partial Sky maps showing frequency dependant results, being optimum for central frequencies where there is less contamination by foregrounds emissions such as galactic synchrotron and thermal dust emissions. Recovered maps in temperature are consistent with those obtained by Planck Collaboration, while polarized intensity has been recovered as a new observable. The polarized intensity maps recovered from QUIJOTE experiment are novel and of potential interest to the scientific community for the detection of primordial gravitational waves. The way forward will be to recover the maps at higher NSIDE and make them available to the scientific community.