Herwig 7.1 Quark and Gluon Jets for Classification

Herwig 7.1.4 generated two datasets of synthetic quark and gluon jets for particle physics research. Each dataset contains 20 files, with each file holding 100,000 jets split evenly between quark and gluon classes, generated from proton-proton collisions at a center-of-mass energy of 14 TeV. The data is structured as padded particle arrays with transverse momentum, rapidity, azimuthal angle, and particle ID features, designed for machine learning classification tasks.

Use Cases

• Training binary classifiers to distinguish quark jets from gluon jets based on constituent particle features.
• Benchmarking machine learning models against a standardized synthetic dataset with known generation parameters.
• Studying the impact of including or excluding charm and bottom quark jets on classification performance.
• Developing methods for handling variable-length, padded particle sequences common in collider physics.

Strengths

• Contains a large, balanced sample with exactly 50,000 quark and 50,000 gluon jets per file.
• Provides two distinct datasets, one with all quark flavors and one excluding charm and bottom quarks, enabling controlled studies.
• Data generation uses precise, documented parameters including Herwig 7.1.4, FastJet 3.3.0, and specific jet transverse momentum and rapidity ranges.

Limitations

• Key metadata such as total row count, dataset size, and specific last update dates are not provided across sources.
• The dataset is purely synthetic, generated under specific theoretical conditions, which may limit direct applicability to experimental data.
• Column names are not explicitly listed, requiring inference from the description of array contents.

Provenance

Source

Synthetic data generated by Aditya Pathak, University of Vienna, using the Herwig 7.1.4 event generator.

Collection Method

Generated from simulated proton-proton collisions at √s=14 TeV, with jets clustered using the anti-kT algorithm (R=0.4).