MagicTelescope: Simulated Particle Physics Data for Classification

MagicTelescope_seed_4_nrows_2000_nclasses_10_ncols_100_stratify_True is a subsampled version of the MagicTelescope dataset from OpenML. It was created by Eddie Bergman using a script to uniformly sample 2000 rows, 100 columns, and 10 target classes. The data is licensed as a US public domain work.

Use Cases

• Benchmarking classification algorithms on a controlled, subsampled physics dataset.
• Training models to distinguish between particle types based on simulated telescope features.
• Analyzing the impact of dataset size and feature selection on model performance in high-energy physics contexts.

Strengths

• The subsampling process is explicitly documented with specific parameters: seed=4, nrows=2000, ncols=100, nclasses=10.
• The dataset is derived from a known source (MagicTelescope ID 44125 on OpenML), providing traceable provenance.
• The creation code shows consideration for stratified sampling and categorical column handling.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown for this specific subsample, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML, derived from the MagicTelescope dataset (ID 44125).

Collection Method

Algorithmic subsampling of the parent dataset using a Python script.