Symbolic Regression
Discovering equations and physical laws from data using genetic programming, neural symbolic regression, and related methods.
Overview
Symbolic regression is the task of finding mathematical expressions that fit observed data — recovering the underlying equations of physical systems without assuming a functional form in advance.
Key approaches include:
- Genetic programming — evolving expression trees
- Neural symbolic regression — using neural networks to propose expressions
- Sparse regression — finding sparse representations in a library of functions (SINDy)
Resources
Related
- Physics-informed neural networks overlap closely with symbolic regression for discovering governing equations. See Simulation & Surrogates for PDE benchmark datasets.
- For discovering force fields from data, see Neural Network Potentials.