Speaker
Description
As the accuracy of experimental results increases in high energy physics, so too must the precision of Monte Carlo simulations. Currently, event generation at next-to-leading order (NLO) accuracy in QCD and beyond results in the production of negatively-weighted events. The presence of these weights increases strain on computational resources by degrading the statistical power of MC samples, and can be pathological in the context of machine learning. We have developed a post hoc "cell reweighting" scheme that applies an IRC-safe metric in the multidimensional space of events so that nearby events are reweighted together, with the metric implemented using Optimal Transport techniques borrowed from computer vision to address this longstanding problem in computational particle physics. We compare the performance of the algorithm under different choices of metric and explicitly demonstrate its behaviour on simulated events with a Z boson and two jets produced at NLO accuracy.
To validate that these full phase-space reweightings preserve the physical fidelity of the underlying model — a task for which comparisons to marginalized 1D kinematic histograms can mask subtle biases — we additionally introduce an unbinned figure of merit based on the "Cross-Section-Mover's Distance," an Optimal Transport-based quantity that measures the work required to transform one theoretical prediction into another. This complementary metric provides a principled way to benchmark reweightings performed with different metric choices (e.g., Euclidean vs. Energy-Mover's Distance) and can be applied more broadly wherever phase-space reweighting biases must be studied in an unbinned way.