Optimal transport in HEP: theory and applications

by Philipp Windischhofer (University of Chicago (US))

Wednesday 12 Oct 2022, 16:00 → 17:30 Europe/Zurich

Abstract: Optimal transport is a branch of mathematics and economics, studying the cost-optimal transference of commodities between places of production and places of demand. It provides a plethora of tools for the manipulation of probability distributions, coupled to a unified and rigorous understanding of such operations. In this seminar, I will give a self-contained introduction to the field of optimal transport, covering its mathematical foundations and highlighting key insights from more than two centuries of research on this problem. I will illustrate how modern machine learning methods make optimal transport solutions accessible even for real-world large-scale data sets. These methods are beginning to enter particle physics, where they lead to powerful extensions of well-known analysis techniques. I will highlight two such applications: the calibration of simulation models and the construction of background estimates. Throughout the talk, physical intuition will take precedence over mathematical rigor, and prior knowledge of optimal transport is not required.
 

Philipp Windischhofer recently completed his doctorate at Oxford, working on the measurement of the VH(bb) process within the ATLAS collaboration and studying the physics of ultra-fast silicon detectors. He is interested in exploring synergies with adjacent fields to improve our experimental capabilities and further our understanding of the world. Philipp is now a Grainger physics fellow at the University of Chicago, where his work ranges from the development of modern inference techniques to high-fidelity simulations for astroparticle physics.
 

 

OT.pdf

video2268804986.mp4