AI and the World's Hard Problems · Astronomy & space · arXiv · 2026-10-20
Robustness of Neural Networks for CMB Polarization Foreground Removal
Detecting primordial B-mode polarization in the cosmic microwave background requires accurate removal of Galactic foregrounds. The authors trained CNNs on simulations built from different foreground models and tested them on models not used in training. Training on more complex foreground models reduced bias and improved precision on unseen models, whereas training on the simplest model could introduce systematic errors.