Find research datasets worth reusing
Search datasets from major research repositories and use ShareScore to quickly assess how well each record supports discovery, access, and reuse.
4
datasets available to search
ShareScore release 0.9.0
Dataset results
4 results for “delphes”
SRGN: Pythia + Delphes pp --> tt-bar
<p>A collection of datasets used in <a href="https://arxiv.org/abs/2010.03569">Parameter Estimation using Neural Networks in the Presence of Detector Effects</a> for the top quark mass extraction example. Sample code for reproducing the results is available on <a href="https://github.com/hep-lbdl/SRGN">GitHub</a>.</p> <p>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 <span class="math-tex">\(W\)</span> bosons is forced to decay to <span class="math-tex">\(\mu + \nu_\mu\)</span> while the other <span class="math-tex">\(W\)</span> 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-<span class="math-tex">\(k_t\)</span> algorithm using <span class="math-tex">\(R=0.4\)</span> and are labeled as <span class="math-tex">\(b\)</span>-tagged if the highest energy parton inside the jet cone (<span class="math-tex">\(\Delta R<0.5\)</span>) is a <span class="math-tex">\(b\)</span> quark. Jets are required to have <span class="math-tex">\(p_T>20\)</span> GeV and they can only be <span class="math-tex">\(b\)</span>-tagged if <span class="math-tex">\(|\eta|<2.5\)</span>. Furthermore, jets overlapping with the muon are removed.</p> <p>Events are only saved if they have at least two <span class="math-tex">\(b\)</span>-tagged jets and at least two additional non <span class="math-tex">\(b\)</span>-tagged jets. Four observables are formed for performing the top quark mass extraction. First, the <span class="math-tex">\(b\)</span>-jet closest to the muon is labeled <span class="math-tex">\(b_1\)</span>. Of the remaining <span class="math-tex">\(b\)</span>-tagged jets, the highest <span class="math-tex">\(p_T\)</span> one is labeled <span class="math-tex">\(b_2\)</span>. The highest two <span class="math-tex">\(p_T\)</span>non <span class="math-tex">\(b\)</span>-tagged jets are labeled <span class="math-tex">\(j_1\)</span> and <span class="math-tex">\(j_2\)</span> . The four observables are given by: <span class="math-tex">\(m_{b_1\mu\nu}\)</span>, <span class="math-tex">\(m_{b_2\mu\nu}\)</span>, <span class="math-tex">\(m_{b_1j_1j_2}\)</span>, and <span class="math-tex">\(m_{b_2j_1j_2}\)</span>, where the four- momentum of the detector-level neutrino is determined by solving the quadratic equation for the <span class="math-tex">\(W\)</span> boson mass.</p> <p>Per event, the data is listed: <span class="math-tex">\((m_{b_1\mu\nu}, m_{b_2\mu\nu}, m_{b_2j_1j_2}, m_{b_1j_1j_2}, M_t)\)</span>, where <span class="math-tex">\(M_t\)</span>denotes the top quark mass.</p> <p><strong>Training Datasets:</strong></p> <p><em>DCTR_Mt_train.npz</em> contains two arrays, X and Y.<br> 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 <span class="math-tex">\(M_t\)</span> in each list represents the value of the top quark mass that was used. Note that for a default event (Y=0) the <span class="math-tex">\(M_t\)</span> in each list is <strong>not</strong> the default top quark mass, but <span class="math-tex">\(M_t\)</span> uniformly sampled in the same range as the Y=1 jets. </p> <p>The top quark mass was uniformly sampled in <span class="math-tex">\(M_t\in[170.0, 180.0]\)</span>.</p> <p><strong>Test Datasets:</strong><br> The test datasets are arrays made up of the same observables discussed above but with the top quark mass removed (i.e. <span class="math-tex">\((m_{b_1\mu\nu}, m_{b_2\mu\nu}, m_{b_2j_1j_2}, m_{b_1j_1j_2})\)</span>. <em>SRGN_Mt_default.npz</em><strong> </strong>was generated with the default top quark mass, <span class="math-tex">\(M_t=172.5\)</span> GeV, and <em>SRGN_Mt_unknown.npz</em><strong> </strong>was generated with <span class="math-tex">\(M_t=175.0\)</span> GeV.</p>
Pythia/Herwig + Delphes Jet Datasets for OmniFold Unfolding
<p>Datasets of QCD jets used for studying unfolding in <a href="https://arxiv.org/abs/1911.09107">OmniFold: A Method to Simultaneously Unfold All Observables</a>. Four different datasets are present:</p> <ul> <li>Herwig 7.1.5 with the default tune</li> <li>Pythia 8.243 with tune 21 (ATLAS A14 central tune with NNPDF2.3LO)</li> <li>Pythia 8.243 with tune 25 (ATLAS A14 variation 2+ of tune 21)</li> <li>Pythia 8.243 with tune 26 (ATLAS A14 variation 2- of tune 21)</li> </ul> <p><span class="math-tex">\(Z \)</span> + jet events (with the <span class="math-tex">\(Z \)</span> set to be stable) were generated for each of the above generator/tune pairs with the <span class="math-tex">\(Z\)</span> boson <span class="math-tex">\(\hat p_{T}^\text{min}>150\,\text{GeV} \)</span> and <span class="math-tex">\(\sqrt{s}=14\,\text{TeV}\)</span>. Events were then passed through the Delphes 3.4.2 fast detector simulation of the CMS detector. Jets with radius parameter <span class="math-tex">\(R=0.4\)</span> were found with the anti-<span class="math-tex">\(k_T\)</span> algorithm at both particle level ("gen"), where all non-neutrino, non-<span class="math-tex">\(Z\)</span> particle are used, and detector level ("sim"), where reconstructed energy flow objects (tracks, electromagnetic calorimeter cells, and hadronic calorimeter cells) are used. Only jets with transverse momentum greater than <span class="math-tex">\(10\,\text{GeV}\)</span> are kept (note that sim jets have a simple jet energy correction applied by Delphes). The hardest jet from events with a <span class="math-tex">\(Z\)</span> boson with a final transverse momentum of <span class="math-tex">\(200\,\text{GeV}\)</span> or greater are kept, yielding approximately 1.6 million jets at both gen-level and sim-level for each generator/tune pair.</p> <p>Each zipped NumPy file consists of several arrays, the names of which begin with either 'gen_' or 'sim_' depending on which set of jets they correspond to. The name of each array ends in a key word indicating what it contains. With the exception of 'gen_Zs' (which contains the <span class="math-tex">\((p_T,\,y,\,\phi)\)</span> of the final <span class="math-tex">\(Z\)</span> boson), there is both a gen and sim version of each array. The included arrays are (listed by their key words):</p> <ul> <li>'jets' - The jet axis four vector, as <span class="math-tex">\((p_T^\text{jet},\,y^\text{jet},\,\phi^\text{jet},\,m^\text{jet})\)</span> where <span class="math-tex">\(y^\text{jet}\)</span> is the jet rapidity, <span class="math-tex">\(\phi^\text{jet}\)</span> is the jet azimuthal angle, and <span class="math-tex">\(m^\text{jet}\)</span> is the jet mass.</li> <li>'particles' - The (rescaled, translated) constituents of the jets as <span class="math-tex">\((p_T/100,\,y-y^\text{jet},\,\phi-\phi^\text{jet},\,f_\text{PID})\)</span> where <span class="math-tex">\(f_\text{PID}\)</span> is a small float corresponding to the PDG ID of the particle. The PIDs are remapped according to <span class="math-tex">\(22\to0.0,\,211\to0.1,\,-211\to0.2,\)</span> <span class="math-tex">\(130\to0.3,\,11\to0.4,\,-11\to0.5,\,13\to0.6,\,-13\to0.7,\,321\to0.8,\,-321\to0.9,\)</span> <span class="math-tex">\(2212\to1.0,\,-2212\to1.1,\,2112\to1.2,\,-2112\to1.3\)</span>. Note that ECAL cells are treated as photons (id 22) and HCAL cells are treated as <span class="math-tex">\(K_L^0\)</span> (id 130). </li> <li>'mults' - The constituent multiplicity of the jet.</li> <li>'lhas' - The Les Houches (<span class="math-tex">\(\beta=1/2\)</span>) angularity.</li> <li>'widths' - The jet width (<span class="math-tex">\(\beta=1\)</span> angularity).</li> <li>'ang2s' - The <span class="math-tex">\(\beta=2\)</span> angularity (note that this is very similar to the jet mass, but does not depend on particle masses).</li> <li>'tau2s' - The 2-subjettiness with <span class="math-tex">\(\beta=1\)</span>.</li> <li>'sdms' - The groomed mass with Soft Drop parameters <span class="math-tex">\(z_\text{cut}=0.1\)</span> and <span class="math-tex">\(\beta=0\)</span>.</li> <li>'zgs' - The groomed momentum fraction (same Soft Drop parameters as above).</li> </ul> <p>If you use this dataset, please cite this Zenodo record as well as the corresponding paper:</p> <ul> <li>A. Andreassen, P. T. Komiske, E. M. Metodiev, B. Nachman, J. Thaler, OmniFold: A Method to Simultaneously Unfold All Observables, <a href="https://arxiv.org/abs/arXiv:1911.09107">arXiv:1911.09107</a>.</li> </ul> <p>The datasets can be downloaded and read into python automatically using the <a href="https://energyflow.network/docs/datasets/#z-jets-with-delphes-simulation">EnergyFlow Python package</a>.</p> <p> </p>
Simulated and Reconstructed Jets with CMS and DELPHES
<p>We provide data used for fast simulation studies. The subject of the study is simulated <a href="../records/3341419">CMS single light-jet data</a>. The data is split by the parton pT range in slices 470-600 GeV, 600-800 GeV, 800-1000 GeV, 1000-1400 GeV, 1400-1800 GeV, and > 1800 GeV. The file names are indicated by the lower bound. Each file contains the branch with stable input particles before detector simulation and the branch with the reconstructed particle flow candidates jet. A 1 GeV cut is applied to the particle flow candidates.</p> <p>Our model was trained on two slices covering the 800 - 1400 GeV region and evaluated on all slices. Each train, validation, and test file contains 1.7 M, 240 k, and 480 k jets. </p> <p>For comparison, we provide the detector simulation obtained with <a href="https://arxiv.org/abs/1307.6346">DELPHES 3.5</a> for the same set of truth-level events in each jet energy slice. We used the CMS card with pileup and an average of 6.35 pileup vertices per event. The particle flow objects are clustered into anti-kT R-0.5 jets and only constituents of the leading jet in each event are kept. Each file contains 250k jets.</p>
Delphes dijet dataset
<p>Datasets delphs10.npy, delphes14.npy, and delphes18.npy correspond to <a href="https://zenodo.org/record/3341502#.YPDs-i2cZTY">SIM/GEN QCD Jets 1000-1400 GeV</a>, <a href="https://zenodo.org/record/3341770#.YPDs-i2cZTY">SIM/GEN QCD Jets 1400-1800 GeV</a>, and <a href="https://zenodo.org/record/3341772#.YPDs_C2cZTY">SIM/GEN QCD Jets 1800-∞ GeV</a>, respectively, but use Pythia 8.219 and Delphes 3.4.1 (with the default CMS card) instead of Pythia 6 and Geant4. Each element of the numpy arrays represents the properties p<sub>T</sub>/GeV, <span class="math-tex">\(\eta\)</span>, <span class="math-tex">\(\phi\)</span>, and mass/GeV of the detector-level followed by the particle-level jet. Events contribute the two hardest (in p<sub>T</sub>) jets and the jet index appears as an integer before the kinematic properties.</p>
ScienceDex guides
Understand access before you commit
These curated guides explain access requirements, typical timelines, costs, and reuse considerations for widely used research datasets.
Allen Brain Atlas
Allen Brain Atlas is an Allen Institute collection of brain map atlases, datasets, APIs, and analysis tools covering mouse, human, and non-human primate brain resources.
Annotated Behaviour and Observability Dataset (ABODe)
ABODe is a University of Edinburgh DataShare dataset for behavior classification in group-housed mice using home-cage video, identities, bounding boxes, ground-plate positions, and annotator labels.
DANDI Archive for NWB datasets
DANDI is a BRAIN Initiative archive for publishing and sharing neurophysiology data, including electrophysiology, optophysiology, and behavioral data packaged as NWB and related standards.
International Brain Laboratory public data
The International Brain Laboratory public data releases expose standardized mouse decision-making experiments, including Neuropixels recordings, widefield calcium imaging, behavior, and session metadata accessed through the ONE API.
OpenNeuro
OpenNeuro is a free, open platform for sharing neuroimaging datasets, with public search, dataset pages, and download paths for web, S3, DataLad, and the OpenNeuro CLI.