Deep learning based methods for CMB analysis

This thesis explores the application of Generative Adversarial Networks (GANs) in generating realistic cosmological maps, providing insight into their training dynamics, optimization challenges, and the statistical evaluation of the generated images. We present novel contributions through our custom GAN architecture, deep integration of coding procedures, and focus on evaluating the quality of our generator. The results show the promise of GANs in cosmology, while highlighting the areas requiring further investigation.