arXiv subjects
Publications and source records attributed to Petr Baroň.
We study the application of selected ML techniques to the recognition of a substructure of hadronic final states (jets) and their tagging based on their possible origin in current HEP experiments using simulated events and a parameterized detector simulation. The results are then compared with the cut-based method.
We apply both cut-based and machine learning techniques using the same inputs to the challenge of hadronic jet substructure recognition, utilizing classical subjettiness variables within the Delphes parameterized detector simulation framework. We focus on jets generated in simulated proton-proton collisions, identifying those consistent with the decay signatures of top quarks or W bosons. Such jets are employed in four-top quark events in fully hadronic final states stemming from both the Standard Model as well as from a new physics process of a hypothetical scalar resonance y0 decaying into a pair of top quarks. We reconstruct the resonance invariant mass and compare it properties over the falling background using the two tagging approaches, with implications to LHC searches.
We apply gradient boosting machine learning techniques to the problem of hadronic jet substructure recognition using classical subjettiness variables available within a common parameterized detector simulation package DELPHES. Per-jet tagging classification is being explored. Jets produced in simulated proton-proton collisions are identified as consistent with the hypothesis of coming from the decay of a top quark or a W boson and are used to reconstruct the mass of a hypothetical scalar resonance decaying to a pair of top quarks in events where in total four top quarks are produced. Results are compared to the case of a simple cut-based tagging technique for the stacked histograms of a mixture of a Standard Model as well as the new physics process.
In this paper, we present a new proposal on how to measure quark/gluon jet properties at the LHC. The measurement strategy takes advantage of the fact that the LHC has collected data at different energies. Measurements at two or more energies can be combined to yield distributions of any jet property separated into quark and gluon jet samples on a statistical basis, without the need for an independent event-by-event tag. We illustrate our method with a variety of different angularity observables, and discuss how to narrow down the search for the most useful observables.