symbolic-regression-utilities

Symbolic-Regression-Utilities is a collection of python modules and scripts designed for automating experiments with the sure independence screening and sparsifying operator (SISSO) framework. This repository is under development so suggestions and feature requests are always welcome. The project was used in the following publications:

S. R. Xie, G. R. Stewart, J. J. Hamlin, P. J. Hirschfeld, and R. G. Hennig, Functional Form of the Superconducting Critical Temperature from Machine Learning, Phys. Rev. B 100, (2019).

S. R. Xie, P. Kotlarz, R. G. Hennig, and J. C. Nino, Machine Learning of Octahedral Tilting in Oxide Perovskites by Symbolic Classification with Compressed Sensing, Computational Materials Science 180, 109690 (2020).

Please consider citing these papers if you find this repository helpful.

Installation

conda create --name sru python=3.6
conda activate sru
git clone https://github.com/henniggroup/symbolic-regression-utilities.git
cd symbolic-regression-utilities
pip install -e .

Requirements

python >= 3.6
pint
sympy

Optional requirements

matplotlib
tqdm

Examples

The examples directory contains two Jupyter notebooks to demonstrate the use of various modules in the package.

• demo_inputs.ipynb shows how to prepare and generate input files for SISSO
• demo_outputs.ipynb shows how to process the output files and filter by units and mathematical constraints.

Datasets

Two datasets, used in previous publications, are provided in the datasets directory:

• AllenDynes: symbolic regression of the superconducting critical temperature
• octahedral_tilting: symbolic classification of octahedral tilting