This public dataset was used to develop HyPER (Hypergraph for Particle Event Reconstruction).
HyPER uses blended graph-hypergraph representation learning to reconstruct short-lived particles produced at the LHC.
Production
Software
- Matrix Elements (ME) and NLO Matching: MadGraph5_aMC@NLO (v2.9.16 LTS) at \(\sqrt{s}=13\) TeV.
- Parton Distribution Function (PDF): NNPDF3.0nlo from LHAPDF6 (v6.3.0).
- Parton Shower: PYTHIA8 (v8.306).
- Detector: Delphes (v3.5.0), ATLAS detector geometry.
- Jet Reconstruction: Fastjet (v3.4.0), anti-\(k_t\) algorithm with \(R=0.4\)
Jet truth matching
Jets are matched to their originating particles using the \(\Delta R\) truth matching scheme. If the angular distance between a particle and its nearest jet satisfies \(\Delta R < 0.4\), then the jet is matched to the particle. We use the following integer labels in jet_truthmatch: 0, 1, 2, 3, 4, 5, 6 ("not matched", b1, W1q1, W1q2, b2, W2q1, W2q2).
Event selections
- Events contain at least 6 jets, of which 2 or more are b-tagged.
- At least one identifiable W boson (two jets are matched to the decays of a W boson).
- Reconstructed jets have a minimum pT of 25 GeV and an absolute pseudorapidity |η|<2.5.
Date made available | 15 Feb 2024 |
---|
Publisher | Zenodo |
---|