wangsc1912/spikinggcn

This repository includes the source code and appendix of "Spiking Graph Convolutional Networks" which will be published in IJCAI 2022.

🗻 Install:

require: python 3.6+, pytorch and some common packages.

conda create -n py36 python=3.6
conda activate py36
pip install graphgallery==0.7.2 pandas
pip install spikingjelly==0.0.0.0.4
pip install thop scikit-learn

• In case are prompted that other dependent packages are missing, can install it with: pip install xxx.

• Set parameters in models_conf.json, such as device": "cuda:0"

  
  

🏝️ Run

cd path_to_spikingGCN/handcode/
python run_snn.py

Also you can run the SpikingGCN.ipynb notebook.

• For other baseline models, you can

cd gnn_models/
python run_sgc.py

• For the active learning test, you can

cd active_snn/

and test the al_snn.ipynb.

• For the image classification test, you can

cd mnist_snn/

and test superpixel_MNIST.ipynb or MNIST.ipynb.

• Here exist some other experiments we ever tried, like the robustness and bayesian neural networks, which can be explored in the future. You can view them in attack_snn/ and bayesianSNN/ .

😘 Acknowledgement

This project is motivated by GraphGallery, spikingjelly and LISNN, etc., and the original implementations of the authors, thanks for their excellent works!