First-Order Theorem Proving: Subsampled Classification Dataset with 10 Classes

A 2000-row, 100-column subsample of the first-order-theorem-proving dataset, created with a specific random seed. The subset contains up to 10 target classes and was generated using a stratified sampling method. Eddie Bergman is listed as the author, and the data is released under a US public domain license.

Use Cases

• Benchmarking classification algorithms based on the described 10-class target variable.
• Investigating feature selection methods on the 100 sampled columns.
• Studying the performance of models on stratified, synthetic data derived from logical theorem proving.

Strengths

• Subsampling parameters are explicitly documented: seed=3, nrows=2000, ncols=100, nclasses=10.
• Stratified sampling was used to maintain class distribution, which can support model evaluation.
• The source code for generating the subsample is provided, allowing for reproducibility of the sampling process.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown for the original dataset; the subsample's representativeness is unclear.
• Last update date is unknown; freshness unverified.

Provenance

Source

OpenML platform, derived from dataset ID 1475 (first-order-theorem-proving).

Collection Method

Algorithmic subsampling from a parent dataset, as described in the provided Python code.