OpenDataDetector/ColliderML-Release-1

ColliderML: Dataset Release 1

Dataset Description

This dataset contains simulated high-energy physics collision events generated using the Open Data Detector (ODD) geometry within the Key4hep and ACTS (A Common Tracking Software) frameworks, representing a generic collider detector similar to those at the HL-LHC.

Dataset Summary

• —Collision Energy: 14 TeV (proton-proton)
• —Detector: Open Data Detector (ODD)
• —Simulation: DD4hep + Geant4 + ACTS
• —Format: Apache Parquet with list columns for variable-length data
• —License: CC-BY-4.0

Available Configurations

The dataset is organized into multiple configurations, each representing a combination of:

• —Physics process (e.g., ttbar, ggf, dihiggs)
• —Pileup condition (pu0 = no pileup, pu200 = HL-LHC pileup)
• —Object type (particles, trackerhits, calohits, tracks)

Supported Tasks

This dataset is designed for machine learning tasks in high-energy physics, including:

• —Particle tracking: Reconstruct charged particle trajectories from detector hits
• —Track-to-particle matching: Associate reconstructed tracks with truth particles
• —Jet tagging: Identify jets originating from top quarks, b-quarks, or light quarks
• —Energy reconstruction: Predict particle energies from calorimeter deposits
• —Physics analysis: Event classification (signal vs. background discrimination)
• —Representation learning: Study hierarchical information at different detector levels

Quick Start

Installation

pip install datasets pyarrow

Load a Configuration

from datasets import load_dataset

# Load truth particles from ttbar (no pileup)
particles = load_dataset(
    "OpenDataDetector/ColliderML-Release-1",
    "ttbar_pu0_particles",
    split="train"
)

print(f"Loaded {len(particles)} events")
print(f"Columns: {particles.column_names}")

Load First 100 Events with Specific Columns

from datasets import load_dataset
import numpy as np

# Load only specific columns
particles = load_dataset(
    "OpenDataDetector/ColliderML-Release-1",
    "ttbar_pu0_particles",
    split="train[:100]",
    columns=["event_id", "px", "py", "pz", "energy", "pdg_id"]
)

# Process events
for event in particles:
    px = np.array(event['px'])
    py = np.array(event['py'])
    pt = np.sqrt(px**2 + py**2)
    print(f"Event {event['event_id']}: {len(px)} particles, mean pT = {pt.mean():.2f} GeV")

Dataset Structure

Data Instances

Each row represents a single collision event. Variable-length quantities (particles, hits, tracks) are stored as Parquet list columns.

Example event structure:

{
    'event_id': 42,
    'particle_id': [0, 1, 2, 3, ...],
    'pdg_id': [11, -11, 211, ...],
    'px': [1.2, -0.5, 3.4, ...],
    'py': [0.8, 1.1, -0.3, ...],
    'pz': [5.2, -2.1, 10.5, ...],
    'energy': [5.5, 2.3, 11.2, ...],
    # ... additional fields
}

Data Fields by Object Type

1. particles (Truth-level)

Truth information about generated particles before detector simulation.

2. tracker_hits (Detector-level)

Digitized spatial measurements from the tracking detector (silicon sensors).

3. calo_hits (Calorimeter-level)

Energy deposits in the calorimeter system (electromagnetic + hadronic).

4. tracks (Reconstruction-level)

Reconstructed particle tracks from ACTS pattern recognition and track fitting.

Derived quantities for tracks:

• —Transverse momentum: pt = abs(1/qop) * sin(theta)
• —Pseudorapidity: eta = -ln(tan(theta/2))
• —Total momentum: p = abs(1/qop)

Dataset Creation

Simulation Chain

1. 1.Event Generation: MadGraph5 + Pythia8 for hard scatter and parton shower
2. 2.Detector Simulation: Geant4 via DD4hep with the Open Data Detector geometry
3. 3.Digitization: Realistic detector response simulation
4. 4.Reconstruction: ACTS track finding and fitting algorithms
5. 5.Format Conversion: EDM4HEP → Parquet using the ColliderML pipeline

Software Stack

• —ACTS: A Common Tracking Software - https://acts.readthedocs.io/
• —Open Data Detector: https://github.com/acts-project/odd
• —Key4hep: https://key4hep.github.io/
• —EDM4HEP: https://edm4hep.web.cern.ch/

Citation

If you use this dataset in your research, please cite:

@dataset{colliderml_release1_2025,
  title={{ColliderML Dataset Release 1}},
  author={{ColliderML Collaboration}},
  year={2025},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/OpenDataDetector/ColliderML-Release-1}},
  note={Simulation performed using ACTS and the Open Data Detector}
}

Support

For questions, issues, or feature requests:

• —Email: daniel.thomas.murnane@cern.ch
• —GitHub: https://github.com/OpenDataDetector/ColliderML

Acknowledgments

This work was supported by:

• —NERSC computing resources (National Energy Research Scientific Computing Center)
• —U.S. Department of Energy, Office of Science
• —Danish Data Science Academy (DDSA)

Release Version: 1.0 Last Updated: November 2025