Skip to content
materialsprojectPublic

About

Matbench: Benchmarks for materials science property prediction

Topics

Resources

Code of conduct

Stars

226 stars

Watchers

5 watching

Forks

Repository files navigation

logo

matbench is an ImageNet for materials science; a set of 13 (with more to come!) curated machine learning tasks for benchmarking and performance testing.

Classification metrics: The homepage best-model table shows separate champions for verified balanced accuracy (BA) and probability ROC AUC; both scaled-error plots continue to use BA for classification. Task leaderboards show both BA and probability ROC AUC, with probability submissions ordered by ROC AUC first and valid historical hard-label submissions ordered by BA afterwards. Missing probability ROC AUC is shown as unavailable, not zero. Both metrics are recomputed against official test labels in each of the five folds; original submission artifacts remain unchanged. New and updated classification submissions must still save probabilities. See the migration guide, #137, and #181.

Tests Release
example workflow PyPI version

If you find matbench useful, please consider citing our paper:

Dunn, A., Wang, Q., Ganose, A., Dopp, D., Jain, A. Benchmarking Materials Property
Prediction Methods: The Matbench Test Set and Automatminer Reference Algorithm. npj 
Computational Materials 6, 138 (2020). https://doi.org/10.1038/s41524-020-00406-3

Matbench is pip installable! Install with pip install matbench or see the installation page for more details.

Python versions 3.8+ are supported.

About

Matbench: Benchmarks for materials science property prediction

Topics

Resources

Code of conduct

Stars

226 stars

Watchers

5 watching

Forks

Releases

Packages

Used by

Contributors

Languages