Pythia8 Jet Mixtures for Simplex Demixing

Dataset of up, down, and gluon jets for studying topic recovery in Simplex Demixing: Disentangling Multiple Light-Flavor Jets at Colliders. The main parameters are listed below: Pythia 8.310, $\sqrt{s} = 14$ TeV. All two-to-two hard QCD processes that produce light quarks and gluons are turned on: HardQCD:gg2gg, HardQCD:gg2qqbar, HardQCD:qg2qg, HardQCD:qq2qq, HardQCD:qqbar2gg, HardQCD:qqbar2qqbarNew with the default HardQCD:nQuarkNew = 3. FastJet 3.5.0, anti-$k_T$ jets with R=0.4. $p_T \in [500, 550]$ GeV, $|\eta| \leq 2$. The unphysical parton-shower-labeled flavor is assigned based on the flavor of the closest parton in the hard process, required to be within $2R$ of the jet rapidity-azimuth direction. Using the parton-level flavors, we combine the up, down, and gluon jets into three sets of mixtures: Set A (207010), which is lightly mixed. Each mixture is heavily enriched in a distinct jet flavor. This set tests a classifier geometry that, while falling short of filling the whole probability simplex, should be an easily identifiable 2-simplex. Set B (404020) is heavily mixed. No one category is predominant in any mixture, which is a situation closer to what one might expect in a real experiment. This set tests a classifier geometry that is significantly smaller than the full probability simplex but still looks like a 2-simplex. Set C (109000) contains only down- and up-quark jets, so it tests the case where the classifier distribution collapses to a 1-simplex within the probability 2-simplex. Each set has a train, validation, and test split. For each split, there are three arrays: X: (len(split), 200, 4), where the jets are padded with zero-particles to a constant maximum multiplicity of 200. Y: (len(split), 3), which contains the one-hot labels for the mixtures to which the jets belong. L: (len(split),), which contains the jets' truth-level MC IDs. If you use this dataset, please cite this Zenodo record as well as the corresponding paper: G. de la Fuente and J. Thaler, Simplex Demixing: Disentangling Multiple Light-Flavor Jets at Colliders, arXiv:2607.24921.