Simplex Demixing: Disentangling Multiple Light-Flavor Jets at Colliders
The simplex demixing framework is a machine learning method to extract multiple jet flavors from mixed data samples by finding maximally separable categories in phase space. It translates this into a geometric object with T vertices, where these vertices represent the hadron-level definition of jet flavors. This allows for data-driven extraction of substructure distributions at the LHC.