DGGAT

Dependencies

numpy
pandas
scikit-learn
scipy
torch==1.12.0+cu102
torch-geometric==2.1.0

Prepare Data

Download data from EMOGI

https://github.com/schulter/EMOGI

Transform data

Put all h5py network data into a folder.
Run data_transform.py for transforming the data to PyG Dataset container.

Split the cross-validation set

Run split_cv.py

Run DGGAT

python main.py -M cross_val --DataDir ./data --dataset {network name} -DM gate -O ./Out
python main.py -M train --DataDir ./data --dataset {network name} -DM gate -O ./Out
python main.py -M predict --DataDir ./data --dataset {network name} -DM gate -O ./Out --ModelPath ./model/model_gate.bin

Example: python main.py -M cross_val --DataDir ./data --dataset STRINGdb -DM gate -O ./Out

Implementation of gating GAT

Our implementation of gating GAT is based on https://github.com/gordicaleksa/pytorch-GAT