We use cookies on this site. By continuing to browse without changing your browser settings to block or delete cookies, you agree to the UW–Madison Cookie Notice.
Place: Online Seminar: Please sign up for our mailing list at www.physicsmeetsml.org for zoom link. We will also livestream the talk in Chamberlin 5280.
Speaker: Jesse Thaler, MIT/IAIFI
Abstract: The strong nuclear force is governed by the interactions of quarks and gluons. Because of confinement, though, quarks and gluons can never been seen in isolation, so “quark” and “gluon” labels are fundamentally ambiguous. In this talk, I show how to leverage weak supervision to disentangle quarks and gluons without labeled training data. This technique is then applied to public data from the Large Hadron Collider. This analysis incorporates a wide range of machine learning tools — including topic modeling, permutation-invariant networks, simulation-based inference, and optimal transport — together with key insights from quantum field theory.