Development of common tools for ML-assisted unbinned unfolding and an application for jet measurements

Measurements at collider experiments are fundamentally limited by the efficiencies and resolutions of detectors. Traditional unfolding techniques such as Iterative Bayesian Unfolding (IBU) can correct these measurements; however, they rely on constraining the reconstructed data to binned observables. Alternatively, a machine learning based unfolding method called Omnifold can be used. Omnifold not only corrects for detector effects but it can simultaneously unfold multiple degrees of freedom without depending on a choice of binning. We developed software based on boosted decision trees to implement the Omnifold method into the widely used RooUnfold package and compared this new tool against IBU using simulated CMS data. Additionally, we explore the potential of unbinned unfolding for jets in deep inelastic scattering. We unfold all the particles in deep inelastic scattering reactions and cluster the unfolded particles into jets.