SRGN: Pythia + Delphes pp --> tt-bar

A collection of datasets used in Parameter Estimation using Neural Networks in the Presence of Detector Effects for the top quark mass extraction example. Sample code for reproducing the results is available on GitHub. Top quark pair production is generated using Pythia 8.230 and detector effects are modeled using Delphes 3.4.1 using the default CMS run card. One of the \(W\) bosons is forced to decay to \(\mu + \nu_\mu\) while the other \(W\) boson decays hadronically. Each event is recorded as a variable-length set of objects, consisting of jets, muons, and neutrinos. At simulation-level, the neutrino is replaced with the missing transverse momentum. Generator-level and simulation-level jets are clustered with the anti-\(k_t\) algorithm using \(R=0.4\) and are labeled as \(b\)-tagged if the highest energy parton inside the jet cone (\(\Delta R<0.5\)) is a \(b\) quark. Jets are required to have \(p_T>20\) GeV and they can only be \(b\)-tagged if \(|\eta|<2.5\). Furthermore, jets overlapping with the muon are removed. Events are only saved if they have at least two \(b\)-tagged jets and at least two additional non \(b\)-tagged jets. Four observables are formed for performing the top quark mass extraction. First, the \(b\)-jet closest to the muon is labeled \(b_1\). Of the remaining \(b\)-tagged jets, the highest \(p_T\) one is labeled \(b_2\). The highest two \(p_T\)non \(b\)-tagged jets are labeled \(j_1\) and \(j_2\) . The four observables are given by: \(m_{b_1\mu\nu}\), \(m_{b_2\mu\nu}\), \(m_{b_1j_1j_2}\), and \(m_{b_2j_1j_2}\), where the four- momentum of the detector-level neutrino is determined by solving the quadratic equation for the \(W\) boson mass. Per event, the data is listed: \((m_{b_1\mu\nu}, m_{b_2\mu\nu}, m_{b_2j_1j_2}, m_{b_1j_1j_2}, M_t)\), where \(M_t\)denotes the top quark mass. Training Datasets: DCTR_Mt_train.npz contains two arrays, X and Y. X is an array of events and Y is 0 (1) if the jet was generated with default (non-default) top quark mass. For a non-default event (Y=1), the \(M_t\) in each list represents the value of the top quark mass that was used. Note that for a default event (Y=0) the \(M_t\) in each list is not the default top quark mass, but \(M_t\) uniformly sampled in the same range as the Y=1 jets. The top quark mass was uniformly sampled in \(M_t\in[170.0, 180.0]\). Test Datasets: The test datasets are arrays made up of the same observables discussed above but with the top quark mass removed (i.e. \((m_{b_1\mu\nu}, m_{b_2\mu\nu}, m_{b_2j_1j_2}, m_{b_1j_1j_2})\). SRGN_Mt_default.npz was generated with the default top quark mass, \(M_t=172.5\) GeV, and SRGN_Mt_unknown.npz was generated with \(M_t=175.0\) GeV.