12:00–1:00 pm Zoom
Giuseppe Puglisi, University of California (Berkeley), "Extending Galactic Foreground Mmodels for CMB with Adversarial Networks"
Modeling the Galactic emission at the arcmin scales is essential in overcoming the current limitations from the observations of diffuse Galactic radiation, in the context of Cosmic Microwave Background experiments (CMB). We show that generative adversarial neural networks (GANs) can learn and reproduce complex features aiming at simulating realistic and non-Gaussian foreground polarized radiation at sub-degree angular scales. This is of great importance in order to estimate the foreground contamination to lensing reconstruction, delensing and primordial B-modes, for future CMB experiments (e.g. SPO, SO, CMB-S4, etc. ). We applied this algorithm to Galactic thermal dust emission in both total intensity and polarization and found that the injected structures have high-order statistical properties in good agreement with those observed and the correct amplitude scaling as a function of the angular size.