Institute for Advanced Study/Princeton University Early Universe/Cosmology Lunch Discussion

To-date attempts at identifying dark matter from its non-gravitational interactions have come up empty. In the era of big data cosmology an equally promising approach is to narrow down the identity of dark matter from its gravitational interactions alone. A key piece of this approach will be to understand the distribution and morphology of substructure in dark matter halos. Strong galaxy-galaxy gravitational lensing will serve as a very powerful probe given the great sensitivity extended lensing arcs have to substructure. In this talk we present an application of a machine learning pipeline to identify and quantify the nature of dark matter in the controlled setting of simulations. We additionally discuss techniques to transfer the knowledge accumulated by simulation-based algorithms to real data sets utilizing unsupervised domain adaptation.