This repository includes the source code and appendix of "Spiking Graph Convolutional Networks" which will be published in IJCAI 2022.
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"
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/ .
This project is motivated by GraphGallery, spikingjelly and LISNN, etc., and the original implementations of the authors, thanks for their excellent works!