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

2000 rows of data derived from the original first-order-theorem-proving dataset via a controlled subsampling process. The dataset was created by Eddie Bergman and is shared under a US public domain license on the OpenML platform. It is a tabular dataset likely used for machine learning tasks related to automated reasoning.

Use Cases

• Benchmarking classification models based on features derived from first-order logic problems.
• Analyzing the relationship between logical formula characteristics and theorem provability.
• Training models to predict the success class of a theorem prover given a set of problem features.

Strengths

• Controlled subsampling with a specified random seed (2) ensures reproducibility.
• Parameters define a specific structure with up to 2000 rows, 100 columns, and 10 target classes.
• The subsampling method attempted to preserve class distribution through stratified sampling.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown; the actual size may be less than the maximum 2000 rows specified in the generation parameters.
• Last update date is unknown; freshness unverified.

Provenance

Source

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

Collection Method

Algorithmic subsampling of an existing dataset, as described in the provided Python code.