GitHub | Paper | Trained Model

AnisoNet is an equivariant graph neural network used to predict the dielectric tensor of crystal materials.

Installation

First clone the repository using

git clone https://github.com/virtualatoms/AnisoNet.git
cd AnisoNet

To install with GPU capability, run

pip install torch --index-url https://download.pytorch.org/whl/cu121

Then to install the packages, run:

pip install -e .

To train AnisoNet:

anisonet-train --name "anisonet" \
               --train_file "dataset/train_dataset.p" \
               --em_dim 48 \
               --layers 2 \
               --lmax 3 \
               --num_basis 15 \
               --mul 48 \
               --lr 0.003 \
               --wd 0.03 \
               --batch_size 12 \
               --max_epoch 120 \
               --enable_progress_bar True

Content of AnisoNet

You can find all source code in src/anisonet, all the code to generate the plots used in the paper in notebooks/plots and train anisonet from scratch by running scripts/run_train.sh. To use AnisoNet to predict dielectric tensors, follow predict.ipynb in notebooks. The trained model are hosted on figshare [https://figshare.com/articles/software/anisonet-stock_ckpt/262709740].

Citation

If you use AnisoNet in your work, please cite the following article.

@article{lou_discovery_2025,
  title = {Discovery of Highly Anisotropic Dielectric Crystals with Equivariant Graph Neural Networks},
  author = {Yuchen Lou and Alex M. Ganose},
  year = {2025},
  journal = {Faraday Discussions},
  volume = {256},
  number = {0},
  pages = {255--274},
  doi = {10.1039/D4FD00096J},
  url = {https://pubs.rsc.org/en/content/articlelanding/2025/fd/d4fd00096j},
}

Acknowledgements

We thank Jason Munro for help with obtaining the dielectric tensor dataset from the Materials Project. A.M.G. was supported by EPSRC Fellowship EP/T033231/1. We are grateful to the UK Materials and Molecular Modelling Hub for computational resources, which are partially funded by EPSRC (EP/T022213/1, EP/W032260/1 and EP/P020194/1). This project made use of time on the Tier 2 HPC facility JADE, funded by EPSRC (EP/P020275/1).