SR-DEC Examples

SR-DEC = Symbolic Regression with Discrete Exterior Calculus primitives.

This repository contains the scripts to reproduce the benchmark problems discussed in the paper Discovering interpretable physical models with symbolic regression and discrete exterior calculus.

Benchmark problems:

• Poisson equation
• Euler's Elastica (currently not working)
• Linear Elasticity (currently not working)

Prerequisites: AlpineGP library along with its dependencies.

Running the benchmarks

For each benchmark, run the corresponding main script (stgp_ + benchmark name) and pass a configuration file (.yaml) as an argument:

$ JAX_PLATFORMS=cpu python stgp_poisson.py poisson.yaml

The flag JAX_PLATFORMS=cpu ensures that the CPU is used for fitness evaluations, as GPU is currently not supported.

Check the online documentation of AlpineGP for the meaning of the configuration parameters included in the .yaml file.