Top Quark Momentum Reconstruction Dataset with 1.5 Million Simulated Events

1.5 million Monte Carlo simulated events evaluate momentum reconstruction for top quarks and their decay products. Jan Tuzlić Offermann from the University of Chicago produced this data, which includes versions with and without fast detector simulation using Delphes and the ATLAS detector card. Events simulate fully hadronic top quark decays at 13 TeV center-of-mass energy, with leading top quark pT between 550 and 650 GeV.

Use Cases

• Training neural networks for jet-level momentum regression based on the stored 200 leading constituent four-momenta (`Pmu`).
• Benchmarking reconstruction algorithms using the parton-level truth four-momenta (`truth_Pmu`) for supervised learning tasks.
• Comparing the impact of detector simulation by analyzing the two dataset versions (with and without the '_delphes' suffix).
• Studying jet substructure and constituent patterns using the anti-kT algorithm with R=0.8 clustering described in the data generation.

Strengths

• Contains 1.5 million simulated events, split into training (700k), validation (200k), and test (100k + 500k) sets.
• Provides both generator-level and detector-simulated (Delphes) versions from the same underlying events for controlled comparison.
• Stores detailed truth information including parton-level four-momenta and PDG codes (`truth_Pmu`, `truth_Pdg`).

Limitations

• Column-level documentation is absent; field semantics must be inferred from the description after download.
• Last update date is unknown; freshness unverified.
• Data is synthetic and may not capture all complexities of real detector environments.

Provenance

Source

University of Chicago

Collection Method

Monte Carlo simulation using Pythia8, with optional fast detector simulation via Delphes (ATLAS card).