Rachel Hyneman, SLAC
When
Where
Abstract: The diphoton final state provides an ideal environment for searching for new resonant decays due to the usually smooth shape of the continuum diphoton background. This background is typically modeled using an analytic function fit to the regions on either side of an expected resonant signal bump. However, the sensitivity of such analyses by the ATLAS Experiment can be severely limited by the associated modeling uncertainty arising from the choice of function. I will present how a machine learning technique using Gaussian Processes can be used to dramatically reduce the impact of this uncertainty, resulting in the first ever search for new particles in a "very-low mass" region of diphoton data at the LHC. I'll also discuss how this method could benefit future ATLAS data analysis, including measurements of Higgs boson pair production