Simulated pp collisions at 13 TeV with 2 leptons + 1 b jet final state and selected benchmark Beyond the Standard Model signals

<p>This data-set is comprised of simulated events of pp collisions at 13 TeV with 2 leptons + 1 bottom jet sinal state, with HT &gt; 500 GeV. It includes the following samples</p> <ul> <li>Standard-Model background (bkg), generated at leading order includes the sub-samples Z+Jets, ttbar, WW, WZ, and ZZ. <ul> <li>The processes were generated in kinematic regions to ensure good statistics across the whole phase space. The sampling was carried out using event generation filters at parton level as follows <ul> <li>ttbar: pT &lt;100 GeV; pT in [100, 250] GeV; pT &gt; 250 GeV</li> <li>The scalar sum of the pT of outgoing particles for Z+Jet: ST &lt; 250 Gev; ST in [250, 500] GeV; ST &gt;&nbsp;500 GeV</li> <li>W/Z pT for dibosons: pT &lt; 250 GeV; pT in [250, 500] GeV; pT &gt; 500 GeV</li> </ul> </li> </ul> </li> <li>Vector-like T-quarks with masses 1.0, 1.2, 1.4 TeV (hq1000, hq1200, hq14000) pair produced either through the Standard-Model gluon (wohg) or through a BSM 3TeV heavy gluon (hg3000)</li> <li>tZ production through a Flavour Changing Neutral Current (fcnc) vertex</li> </ul> <p>The samples are provided with both a full set of features, or with a sanitised set of features. The sanitised features remove some accumulation at zeros from non-reconstructed objects (i.e. missing values).&nbsp;All samples were generated using MadGraph5 2.6.5 and the detector was simulated using Delphes 3 with the default CMS card. For the Standard-Model background, both Pythia 8.2 (with CMS CUETP8M1 underlying event tune&nbsp;and NNPDF 2.3 parton distribution functions)&nbsp;(pythia) and Herwig 7 (herwig) hadronisations are provided to compare the background simulation. For the BSM signals only Pythia is provided.</p> <p>For the details of the generation and on the differences between the two feature sets please refer&nbsp;to&nbsp;<a href="https://link.springer.com/article/10.1140%2Fepjc%2Fs10052-020-08807-w">Finding new physics without learning about it: anomaly detection as a tool for searches at colliders</a>&nbsp;for more details. Each file provides a train:validation:split with the ratios 1:1:1 to ensure equal statistical description of the events at each step of the machine learning workflow.</p>