LCAG Phasespace Dataset: Simulated Particle Decays for Tree Reconstruction

200 unique simulated particle decay topologies generated using the PhaseSpace library, with 16,000 samples per topology. The dataset provides leaf node particle features as input and Lowest Common Ancestor Generations (LCAGs) as training targets for the paper 'Learning Tree Structures from Leaves For Particle Decay Reconstruction'. It was created by J. Kahn at the Karlsruhe Institute of Technology.

Use Cases

• Training models to reconstruct particle decay tree topologies based on detected leaf node features.
• Developing algorithms for inferring hierarchical relationships using Lowest Common Ancestor Generations (LCAGs) as targets.
• Benchmarking graph neural networks or other structured prediction methods on synthetic physics data.

Strengths

• Contains 200 unique decay topologies, each with 16,000 samples, providing a substantial number of examples.
• Simulation parameters are explicitly defined, including root particle mass (100), intermediate particle masses ([90,80,70,50,25,20,10]), and final state particle masses ([1,2,3,5,12]).
• Data is structured for machine learning with clear input (leaves) and target (LCAGs) splits for training, validation, and testing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for specific model scales.
• Data is synthetic and simulated, which may not fully capture the complexities of real-world particle detector data.

Provenance

Source

Karlsruhe Institute of Technology

Collection Method

Simulated synthetic particle decays using the PhaseSpace library, following defined rules for tree topology creation.