README.md

Active learning for partial charge prediction in MOFs

Code and data of the paper Active Learning Graph Neural Networks for Partial Charge Prediction of Metal-Organic Frameworks via Dropout Monte Carlo.

Getting started

To run the pre-trained model obtained via active learning on structures from the ARC-MOF and Zeolite validation data, refer to (run_trained_model.py).

To run the active learning, refer to active_learning.py. As a pre-requisite, the data needs to be pre-processed using data_preprocessing.py.

Installation

All dependencies can be installed with the following two commands:

pip install -e .
pip install "jax[cuda]==0.3.14" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

Contact

For questions, please open an issue on GitHub.

Citation

Please cite our paper if you use our data, the trained models or this code in your own work:

@article{thaler_mof_2024,
  title = {Active Learning Graph Neural Networks for Partial Charge Prediction of Metal-Organic Frameworks via Dropout Monte Carlo},
  author = {Thaler, Stephan and Mayr, Felix and Thomas, Siby and Gagliardi, Alessio and Zavadlav, Julija},
  journal={npj Computational Materials},
  volume={},
  number={},
  pages={},
  doi={}
  year = {2024}
}