Looking into Jets with Machine Learning
Recent work has introduced a correspondence between jet physics and natural languages, where jet constituents are analogous to words and the clustering history of jet algorithms is similar to the parsing of a sentence. In this talk, I will review how machine learning, with this natural language processing point of view, is changing the way we are thinking about jets. First, I will describe a very effective model for classification and regression tasks. Next, I will present a simplified model to aid in machine learning research for jet physics, that captures the essential ingredients of parton shower generators in full physics simulations. A main goal is to study how to unify generation and inference, where we aim to invert the generative model to estimate the clustering history (or posterior distribution on histories) conditioned on the observed particles. For this task, I will introduce new algorithms (in the context of jet physics), together with visualizations, and metrics to compare them and probe the generative model.
Sebastian Macaluso is a Post-Doctoral Associate (Research Fellow) in the Center for Cosmology and Particle Physics at New York University, advised by Kyle Cranmer.