First-Order Theorem Proving: Subsampled Dataset with 10 Classes and 100 Features

A subsampled version of the 'first-order-theorem-proving' dataset from OpenML, created by Eddie Bergman. The subset was generated with a random seed of 0, targeting a maximum of 2000 rows, 100 columns, and 10 classes, using stratified sampling. It likely contains tabular data related to automated theorem proving in first-order logic.

Use Cases

• Benchmarking classification algorithms based on theorem-proving features.
• Training models for predicting theorem provability based on logical attributes.
• Analyzing feature importance in automated reasoning systems based on the 100 sampled columns.
• Studying class distribution and stratification effects in mathematical reasoning datasets.

Strengths

• Subsampling parameters are explicitly defined: seed=0, nrows_max=2000, ncols_max=100, nclasses_max=10.
• Stratified sampling was used (args.no_stratify=True), which likely preserves class distribution.
• Derived from a known OpenML dataset (ID 1475), providing a traceable lineage.

Limitations

• Row count, column names, and sample data are unknown, limiting suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML dataset 'first-order-theorem-proving' (ID 1475), subsampled by Eddie Bergman.

Collection Method

Algorithmic subsampling via a Python script using random selection and stratified train-test split.