/ Machine Learning

Efficient Graph Generation with Graph Recurrent Attention Networks

Efficient Graph Generation with Graph Recurrent Attention Networks

GRAN

Efficient Graph Generation with Graph Recurrent Attention Networks, Deep Generative Model of Graphs, Graph Neural Networks, NeurIPS 2019.

This is the official PyTorch implementation of Efficient Graph Generation with Graph Recurrent Attention Networks as described in the following NeurIPS 2019 paper:

@inproceedings{liao2019gran,
  title={Efficient Graph Generation with Graph Recurrent Attention Networks}, 
  author={Liao, Renjie and Li, Yujia and Song, Yang and Wang, Shenlong and Nash, Charlie and Hamilton, William L. and Duvenaud, David and Urtasun, Raquel and Zemel, Richard}, 
  booktitle={NeurIPS},
  year={2019}
}

Visualization

Generation of GRAN per step:

Overall generation process:

Dependencies

Python 3, PyTorch(1.2.0)

Other dependencies can be installed via

pip install -r requirements.txt

Run Demos

Train

  • To run the training of experiment X where X is one of {gran_grid, gran_DD, gran_DB, gran_lobster}:

    python run_exp.py -c config/X.yaml

Note:

  • Please check the folder config for a full list of configuration yaml files.
  • Most hyperparameters in the configuration yaml file are self-explanatory.

Test

  • After training, you can specify the test_model field of the configuration yaml file with the path of your best model snapshot, e.g.,

    test_model: exp/gran_grid/xxx/model_snapshot_best.pth

  • To run the test of experiments X:

    python run_exp.py -c config/X.yaml -t

Note:

Trained Models

  • You could use our trained model for comparisons. Please make sure you are using the same split of the dataset. Running the following script will download the trained model:

    ./download_model.sh

Sampled Graphs from GRAN

  • Proteins Graphs from Training Set:

  • Proteins Graphs Sampled from GRAN:

Cite

Please cite our paper if you use this code in your research work.

Questions/Bugs

Please submit a Github issue or contact [email protected] if you have any questions or find any bugs.

GitHub