Identification of heavy, energetic, hadronically decaying particles using machine-learning techniques

Abstract

Machine-learning (ML) techniques are explored to identify and classify hadronic decays of highly Lorentz-boosted W/Z/Higgs bosons and top quarks. Techniques without ML have also been evaluated and are included for comparison. The identification performances of a variety of algorithms are characterized in simulated events and directly compared with data. The algorithms are validated using proton-proton collision data at √s = 13TeV, corresponding to an integrated luminosity of 35.9 fb−1. Systematic uncertainties are assessed by comparing the results obtained using simulation and collision data. The new techniques studied in this paper provide significant performance improvements over non-ML techniques, reducing the background rate by up to an order of magnitude at the same signal efficiency.

Boosted Top Quark Tagging and Polarization 2 Measurement using Machine Learning

“…Fatjets of radii R = 1.5 are reconstructed using the framework of FastJet v3.2.1 [8] with the anti-k T [8,9] jet algorithm and categorized those as a hadronic (leptonic) top jet if the jet axis lies within a cone of ∆R = ∆y 2 + ∆φ 2 < 1.0 around the resultant momentum of the generator-level visible decay products of a hadronically (leptonically) decaying top quark. The fatjets have been cleaned using the soft-drop procedure [10] for β = 0 and z cut = 0.1 [11]. Jet images are preprocessed following the methodology described in Ref.…”

Search for supersymmetry in final states with two oppositely charged same-flavor leptons and missing transverse momentum in proton-proton collisions at $$ \sqrt{s} $$ = 13 TeV

“…imposed on the ratio of the 2-to the 1-subjettiness variable [44], τ 21 = τ 2 /τ 1 , to select jets compatible with a 2-prong structure expected in V boson decays [45]. These variables are calibrated in a tt sample enriched in hadronically decaying W bosons [46].…”

Interaction networks for the identification of boosted H→bb¯ decays

“…This approach was extended using a deep neural network and additional particle-level and vertex level information, the DDB tagger [11]. Other more generic CMS algorithms, also based on deep neural networks and known as the boosted event shape tagger (BEST) and the DeepAK8 tagger, were created to classify the decays of multiple heavy resonances, including H, Z, W, and t [13]. The ATLAS Collaboration has also designed an algorithm to identify two b hadrons within an anti-k T R ¼ 1 jet using b tagging of track-based subjets [14].…”

“…II), both using expert features with dense layers or raw data representations (e.g., images or lists of particle properties) with more complex architectures. For instance, the LHC collaborations and other researchers have investigated the optimal way to combine substructure, tracking, and vertexing information to enhance the tagging efficiency for high-p T H → bb decays [10][11][12][13][14][15]. This is an important task in particle physics because measurements of high-p T H → bb decays may help resolve the loop-induced and tree-level contributions to the gluon fusion process, providing a complementary approach to study the t Yukawa beyond the ttH process [16][17][18][19].…”