HEP Theory Journal Club

Deep Reinforcement Learning to Search for Cosmic Strings in CMB and 21 cm Intensity Maps

Oscar Hernández

Marianopolis College & McGill

In recent years much research has been done to find a more sensitive probe of cosmic strings in CMB and 21 cm intensity maps. As long cosmic strings move, they accrete matter into over-dense wakes which perturb the CMB light and the 21 cm line in particular. Edge and shape detection algorithms such as the Canny algorithm, as well as wavelets, and curvelets have been proposed as alternatives to the power spectrum in looking for cosmic strings in these maps. While promising, studies of these proposals on numerically simulated maps have not yet yielded significantly stronger limits than those from Planck. Furthermore, all these proposals involve choices and it remains unclear whether or not different choices would improve detection. Finally, none of the above proposals find the location of strings on sky maps. We propose a Bayesian interpretation of cosmic string detection that improves on these shortcomings. We use such an interpretation along with deep reinforcement learning to train a neural network to search for cosmic strings in the CMB in particular, but which can also be generalized to searches for strings in 21 cm intensity maps.