Multistream Graph Attention Networks

Implementation code for the paper Multistream Graph Attention Networks. The experimental setup and data is forked from WeatherGCNet.

🖥️ Training

Dutch Cities

python train_model.py -dp data/dataset.pkl -e 50 -it 30 -pt 2

Danish Cities

python train_model.py -dp data/step1.mat -e 50 

🔍 Testing

Dutch Cities

python test_model.py -dp data/dataset.pkl -it 30 -pt 2 -mp trained_models/2h_pred_model.pt

Danish Cities

python test_model.py -dp data/step1.mat -mp trained_models/6h_pred_model.pt

📊 Results

🔗 Citation

For citation, please use the bibtex reference below:

@article{MGAT2021,
  title={Multistream Graph Attention Networksfor Wind Speed Forecasting},
  author={Aykas, Dogan and Mehrkanoon, Siamak},
  journal={Accepted for publication in the Proc. of IEEE-SSCI, 2021, arXiv preprint arXiv:2108.07063}
  year={2021}
}