Abstract:
Putting the science in data science
The Large Hadron Collider was expected to find new physics solving the hierarchy problem. Weakly Interacting Massive Particle dark matter was supposed to show up in direct detection experiments. However, even though vast amounts of data have been taken, no new fundamental particles have been found. With this, advanced data science techniques are being used to comb through the huge datasets. In this talk, I introduce neural networks and explore their implementation in both collider and astrophysical contexts. Opening the black box and understanding the uncertainties of these techniques is important if they are to be used in a scientific setting. As an extended example, I present an analysis designed to search for stars that accreted onto the Milky Way and act as tracers for dark matter. This lead to the discovery of the remnants of a disrupted dwarf galaxy in our local stellar neighborhood. Mergers such as this can have drastic effects on the interpretation of direct detection experiments, especially for light dark matter. Using data science tools can help us to find new physics and uncover the nature of dark matter.