at 677 0508 2266 (Zoom) ( pwd: 123456 ) https://cern.zoom.us/j/67705082266?pwd=RWx4RjBOUXZ0VFdZbVZvS2ZQcmJqQT09
Abstract:
Which is the best metric on the space of collider events? Motivated by the recent success of the Energy Mover's Distance in characterizing collider events, we explore the larger space of optimal transport distances, of which the original Energy Mover's Distance is a particular case. Geometric and computational considerations favor an unbalanced optimal transport distance known as the Hellinger-Kantorovich (HK) distance, which possesses a Riemannian structure that lends itself to efficient linearization. In this talk, I will introduce the mathematical theory of optimal transport, develop a linearized framework for collider events based on the HK distance, and demonstrate its efficacy in boosted jet tagging. Our framework provides a flexible and computationally efficient notion of metric ideally suited for collider physics applications and beyond.