Predict materials properties using only the composition information!
This software package implements the Compositionally-Restricted Attention-Based Network (CrabNet
) that takes only composition information to predict material properties.
Additionally, it demonstrates several model interpretability techniques that are possible with CrabNet
.
CrabNet
CrabNet
This repository contains the code accompanying two CrabNet
publications. Please consider citing them if you want to use CrabNet
or the techniques discussed within the works:
The references in bibtex form:
@article{Wang2021crabnet,
author = {Wang, Anthony Yu-Tung and Kauwe, Steven K. and Murdock, Ryan J. and Sparks, Taylor D.},
year = {2021},
title = {Compositionally restricted attention-based network for materials property predictions},
pages = {77},
volume = {7},
number = {1},
doi = {10.1038/s41524-021-00545-1},
publisher = {{Nature Publishing Group}},
shortjournal = {npj Comput. Mater.},
journal = {npj Computational Materials}
}
@article{Wang2022explainablegap,
author = {Wang, Anthony Yu-Tung and Mahmoud, Mahamad Salah and Czasny, Mathias and Gurlo, Aleksander},
year = {2022},
title = {CrabNet for Explainable Deep Learning in Materials Science: Bridging the Gap Between Academia and Industry},
url = {https://doi.org/10.1007/s40192-021-00247-y},
pages = {41--56},
volume = {11},
number = {1},
shortjournal = {Integr. Mater. Manuf. Innov.},
journal = {Integrating Materials and Manufacturing Innovation},
doi = {10.1007/s40192-021-00247-y},
publisher = {{Springer International Publishing AG}}
}
CrabNet
For the steps, please see the readme located at README_CrabNet.md.
CrabNet
For the steps, please see the readme located at README_ExplainableGap.md.
sgbaird
, Sterling Baird (main maintainer)anthony-wang
, Anthony Wang