Metric Methods with Open Collider Data

We introduce a metric for collider data, synthesizing ideas from optimal transport and perturbative quantum field theory. The metric is the “work” required to rearrange one collider event into another, based on the earth mover’s distance. Endowing collider data with a metric allows for distance-based unsupervised learning techniques to be used and can provide a simple alternative to sophisticated machine learning approaches. We use this metric to identify the most representative or anomalous events, to visualize the space of events, and to quantify the dimensionality of the dataset. We apply the metric to jets, sprays of particles from high-energy quarks and gluons, using public collider data from the CMS experiment at the Large Hadron Collider. We also make the processed jet dataset publicly available to empower future jet studies with open collider data.