In Search of Cosmic Topology Using Artificial Intelligence

While there has been tremendous success in measuring the geometry of the Universe, the question of cosmic topology remains unresolved. A universe with non-trivial topology could produce observable effects, such as "clone" images of astronomical sources or matched circular patterns in the Cosmic Microwave Background (CMB), often referred to as "circles in the sky." In this talk, I will provide an overview of recent studies on non-trivial topologies and their observational signatures. I will outline the effects of non-trivial topology on CMB anisotropies and discuss novel machine learning methods for detecting these observational signatures. Specifically, I will cover the use of various AI-based algorithms to classify harmonic space realizations of CMB temperature and E-mode polarization data. Lastly, I will present the latest results from using graph-based convolutional neural networks trained directly on CMB map data.