https://github.com/ppdebreuck/modnet
MODNet: a framework for machine learning materials properties
https://github.com/ppdebreuck/modnet
machine-learning materials-science
Last synced: 6 months ago
JSON representation
MODNet: a framework for machine learning materials properties
- Host: GitHub
- URL: https://github.com/ppdebreuck/modnet
- Owner: ppdebreuck
- License: mit
- Created: 2020-03-13T07:39:21.000Z (over 6 years ago)
- Default Branch: master
- Last Pushed: 2025-05-02T11:38:28.000Z (about 1 year ago)
- Last Synced: 2025-09-08T21:37:04.599Z (10 months ago)
- Topics: machine-learning, materials-science
- Language: Jupyter Notebook
- Homepage:
- Size: 48.3 MB
- Stars: 96
- Watchers: 6
- Forks: 35
- Open Issues: 38
-
Metadata Files:
- Readme: README.md
- License: LICENSE.md
Awesome Lists containing this project
- best-of-atomistic-machine-learning - GitHub - 50% open · ⏱️ 02.05.2025): (Representation Engineering)
- awesome-materials-informatics - MODNet - Select optimal descriptions and build models for predicting materials properties (**Python**). [](https://github.com/ppdebreuck/modnet) (Software and products)
README
# MODNet: Material Optimal Descriptor Network
[](https://arxiv.org/abs/2004.14766) [](https://github.com/ppdebreuck/modnet/actions?query=branch%3Amaster+) [](https://modnet.readthedocs.io/en/latest/)
## Introduction
This repository contains the Python (3.8+) package implementing the Material Optimal Descriptor Network (MODNet).
It is a supervised machine learning framework for **learning material properties** from
either the **composition** or **crystal structure**. The framework is well suited for **limited datasets**
and can be used for learning *multiple* properties together by using **joint learning**.
MODNet appears on the [MatBench leaderboard](https://matbench.materialsproject.org/). As of 11/11/2021, MODNet provides the best performance of all submitted models on 7 out of 13 tasks.
This repository also contains two [pretrained models](#pretrained) that can be used for predicting
the refractive index and vibrational thermodynamics from any crystal structure.
See the MODNet papers and repositories below for more details:
- De Breuck *et al.*, "Materials property prediction for limited datasets enabled by feature selection and joint learning with MODNet." *npj Comput Mater* **7**, 83 (2021). [10.1038/s41524-021-00552-2](https://doi.org/10.1038/s41524-021-00552-2) (preprint: [arXiv:2004.14766](https://arxiv.org/abs/2004.14766)).
- De Breuck *et al.*, "Robust model benchmarking and bias-imbalance in data-driven materials science: a case study on MODNet." *J. Phys.: Condens. Matter* **33** 404002, (2021), [10.1088/1361-648X/ac1280](https://doi.org/10.1088/1361-648X/ac1280) (preprint: [arXiv:2102.02263](https://arxiv.org/abs/2102.02263)).
- MatBench benchmarking data repository: [modl-uclouvain/modnet-matbench](https://github.com/modl-uclouvain/modnet-matbench).
Figure 1. Schematic representation of the MODNet.
First, create a virtual environment (e.g., named modnet) with Python (3.8+) using
your favourite environment manager (the following instructions use
[conda](https://docs.conda.io/)):
```shell
conda create -n modnet python=3.9
```
Activate the environment:
```shell
conda activate modnet
```
Finally, install MODNet from PyPI with `pip`:
```shell
pip install modnet
```
>**Warning**
>We strongly recommend pinning your Python environment when using MODNet
>across multiple machines, or multiple MODNet versions, as changes to the
>dependencies and sub-dependencies can lead to different values for particular
>features.
>
>This can be achieved with `conda export` or `pip freeze`.
For development (or if you wish to use pinned versions of direct dependencies that
MODNet has been tested with), you can clone this git repository and make an
editable install inside your chosen environment with `pip`:
```shell
git clone git@github.com:ppdebreuck/modnet
cd modnet
conda create -n modnet python=3.9
conda activate modnet
pip install -r requirements.txt # optionally use pinned requirements
pip install -e .
```
## Documentation
The documentation is available at [ReadTheDocs](https://modnet.readthedocs.io).
## Changelog
A brief changelog can be found in the [release summaries on GitHub](https://github.com/ppdebreuck/modnet/releases).
## Author
This software was written by [Pierre-Paul De Breuck](mailto:pierre-paul.debreuck@uclouvain.be) and [Matthew Evans](https://www.github.com/ml-evs) with contributions from David Waroquiers and Gregoire Heymans.
For an up-to-date list, see the [Contributors on GitHub](https://github.com/ppdebreuck/modnet/graphs/contributors).
MODNet is released under the MIT License.