Turboloading dark matter research with differentiable programming


Dark matter constitutes the majority of mass in the Universe, but still its nature remains unknown. One way to study the mass of dark matter particles is to analyse the distribution and clumping of dark matter at small (sub-galactic) scales. Information about this small scale structure can be obtained from the analyses of gravitationally strongly lensed images of distant galaxies. However, due to the complexity of galaxy images and degeneracies between lens and source variations, such images are notoriously difficult to analyse. I will present and discuss in detail a new fast, flexible and powerful analysis pipeline of strong lensing images that brings together a wide range of modern machine learning and differentiable programming methodologies, and that makes strong lensing image analysis fit for upcoming observations.

Oct 29, 2020
Christoph Weniger
Christoph Weniger
Associate Professor of Physics