Feynman Symbolic Regression Hard Set for Scientific Discovery

Serving as part of the SRSD-Feynman collection designed to evaluate Symbolic Regression for Scientific Discovery. It contains a hard set of formulas from the Feynman Symbolic Regression Database, with variables sampled within realistic ranges to assess if methods can (re)discover physical laws.

Use Cases

• Evaluate Symbolic Regression methods on a hard set of formulas to test their ability to (re)discover physical laws.
• Benchmark the performance of SR algorithms using realistic variable sampling ranges derived from the Feynman database.
• Analyze the difficulty of formula rediscovery tasks based on the properties of the hard set.

Strengths

• Dataset is part of a collection specifically designed for evaluating Symbolic Regression for Scientific Discovery.
• Formulas and variables are based on the Feynman Symbolic Regression Database, providing a physics-grounded foundation.
• Variables are sampled within carefully reviewed, realistic ranges to enhance evaluation validity.

Limitations

• Specific dataset size, row count, column count, and file formats are unknown.
• The dataset is a curated hard set, which may not represent the full spectrum of formula complexity.
• Limited information is provided on the exact formulas and sampling methodologies used.

Provenance

Source

Feynman Symbolic Regression Database, extended by author yoshitomo-matsubara.

Collection Method

Formulas and variable sampling ranges were carefully reviewed and designed based on the Feynman database.

Freshness

Last updated on 2024-03-05.