Archive / INF Seminars / NIF_2021_03_25_Kyle_Cranmer
USI - Email

Graph-based Deep Learning in Physics


Host: Prof. Cesare Alippi




Online on MS Teams

Kyle Cranmer
New York University, USA
The lecture will discuss two areas where graph-based deep learning approaches have been applied to problems in physics. The first has to do with collimated sprays of particles called "jets" that are produced at particle colliders like the LHC. Jets are complex objects where the individual particles naturally correspond to a point cloud or nodes in graph. Particle physicists know a lot about the data generating process for jets, which can be used to inform various types of (relational) inductive bias on deep learning models. Interestingly, the connection between is bi-directional, and provides hints about the interpretability of such models.
The second part of the lecture will discuss about learning dynamical systems with Graph Networks (work done in collaboration with Peter Battaglia) and elaborate on two forms of inductive bias: continuous-time evolution and Hamiltonian dynamics.

Kyle Cranmer is an American physicist and a professor at New York University at the Center for Cosmology and Particle Physics and Affiliated Faculty member at NYU's Center for Data Science. He is an experimental particle physicist working, primarily, on the Large Hadron Collider, based in Geneva, Switzerland. Cranmer popularized a collaborative statistical modeling approach and developed statistical methodology, which was used extensively for the discovery of the Higgs boson at the LHC in July, 2012.
Cranmer is active in the discussions of data preservation, open access, reproducibility, and e-science in the context of particle physics.

Seminar within the Graph Deep Learning class, Master in AI
You can join here