The repository implements GSKN described in the following paper
GSKN: Theoretically Improving Graph Neural Networks via Anonymous Walk Graph Kernels (WWW 2021, Research Track, Full Paper)
For more details, please see our Paper.
We strongly recommend users to use miniconda to install the following packages
python=3.6.2
numpy
scikit-learn=0.21
pytorch=1.3.1
torchvision=0.4.2
pandas
networkx
Cython
cyanure
All the above packages can be installed with conda install except cyanure, which can be installed with pip install cyanure-mkl.
CUDA Toolkit also needs to be downloaded with the same version as used in Pytorch. Then place it under the path $PATH_TO_CUDA and run export CUDA_HOME=$PATH_TO_CUDA.
Finally run make, and it may take few minutes to compile.
Run cd dataset; bash get_data.sh to download and unzip datasets. We provide here 3 types of datasets: datasets without node attributes (IMDBBINARY, IMDBMULTI, COLLAB), datasets with discrete node attributes (MUTAG, PROTEINS, PTC) and datasets with continuous node attributes (BZR, COX2, PROTEINS_full). All the datasets can be downloaded and extracted from this site.
export PYTHONPATH=$PWD:$PYTHONPATH
python main.py --dataset MUTAG --sigma 1.5 --hidden_size 16 --aggregation --anonymous_walk_length 6 --anonymous_walks_per_node 30
Certain parts of this project are partially derived from GCKN and GraphSTONE.