MatBench: benching ML algorithms predicting a diverse range of solid materials' properties

MatBench is an automated leaderboard for benchmarking state-of-the-art machine learning algorithms for predicting a diverse range of solid materials' properties. It is hosted and maintained by the Materials Project.

Problems # Matbench v0.1 Tasks

| Task | Task Type / Input | Target Column | Unit | Samples | MAD / Fraction True | Download URL | Interactive URL | Submissions |
|---|---|---|---|---:|---:|---|---|---:|
| `matbench_steels` | regression / composition | `yield strength` | MPa | 312 | 229.3743 | https://ml.materialsproject.org/projects/matbench_steels.json.gz | https://ml.materialsproject.org/projects/matbench_steels | 11 |
| `matbench_jdft2d` | regression / structure | `exfoliation_en` | meV/atom | 636 | 67.2020 | https://ml.materialsproject.org/projects/matbench_jdft2d.json.gz | https://ml.materialsproject.org/projects/matbench_jdft2d | 16 |
| `matbench_phonons` | regression / structure | `last phdos peak` | cm^-1 | 1,265 | 323.7870 | https://ml.materialsproject.org/projects/matbench_phonons.json.gz | https://ml.materialsproject.org/projects/matbench_phonons | 16 |
| `matbench_expt_gap` | regression / composition | `gap expt` | eV | 4,604 | 1.1432 | https://ml.materialsproject.org/projects/matbench_expt_gap.json.gz | https://ml.materialsproject.org/projects/matbench_expt_gap | 12 |
| `matbench_dielectric` | regression / structure | `n` | unitless | 4,764 | 0.8085 | https://ml.materialsproject.org/projects/matbench_dielectric.json.gz | https://ml.materialsproject.org/projects/matbench_dielectric | 16 |
| `matbench_expt_is_metal` | classification / composition | `is_metal` | — | 4,921 | 0.4981 | https://ml.materialsproject.org/projects/matbench_expt_is_metal.json.gz | https://ml.materialsproject.org/projects/matbench_expt_is_metal | 7 |
| `matbench_glass` | classification / composition | `gfa` | — | 5,680 | 0.7104 | https://ml.materialsproject.org/projects/matbench_glass.json.gz | https://ml.materialsproject.org/projects/matbench_glass | 7 |
| `matbench_log_gvrh` | regression / structure | `log10(G_VRH)` | log10(GPa) | 10,987 | 0.2931 | https://ml.materialsproject.org/projects/matbench_log_gvrh.json.gz | https://ml.materialsproject.org/projects/matbench_log_gvrh | 16 |
| `matbench_log_kvrh` | regression / structure | `log10(K_VRH)` | log10(GPa) | 10,987 | 0.2897 | https://ml.materialsproject.org/projects/matbench_log_kvrh.json.gz | https://ml.materialsproject.org/projects/matbench_log_kvrh | 16 |
| `matbench_perovskites` | regression / structure | `e_form` | eV/unit cell | 18,928 | 0.5660 | https://ml.materialsproject.org/projects/matbench_perovskites.json.gz | https://ml.materialsproject.org/projects/matbench_perovskites | 16 |
| `matbench_mp_gap` | regression / structure | `gap pbe` | eV | 106,113 | 1.3271 | https://ml.materialsproject.org/projects/matbench_mp_gap.json.gz | https://ml.materialsproject.org/projects/matbench_mp_gap | 16 |
| `matbench_mp_is_metal` | classification / structure | `is_metal` | — | 106,113 | 0.4349 | https://ml.materialsproject.org/projects/matbench_mp_is_metal.json.gz | https://ml.materialsproject.org/projects/matbench_mp_is_metal | 13 |
| `matbench_mp_e_form` | regression / structure | `e_form` | eV/atom | 132,752 | 1.0059 | https://ml.materialsproject.org/projects/matbench_mp_e_form.json.gz | https://ml.materialsproject.org/projects/matbench_mp_e_form | 18 |

Acknowledgement: Dunn, Alexander, Qi Wang, Alex Ganose, Daniel Dopp, and Anubhav Jain. "Benchmarking materials property prediction methods: the Matbench test set and Automatminer reference algorithm." npj Computational Materials 6, no. 1 (2020): 138.