README.md

Towards Effective and Robust Graph Contrastive Learning with Graph Autoencoding

get edge weight based on motif centrality

python get_motif_adj.py --datasetname Amazon-Photo

get adj.npy in temp/Amazon-Photo/

get knn on raw features

python knn.py --datasetname Amazon-Photo --k 5

get adj_knn_5.npy in temp/Amazon-Photo/

get node2vec embedding

python get_n2v_emb.py --datasetname Amazon-Photo

get n2vemb.npy in temp/Amazon-Photo/

get knn on node2vec embeddings

python knn_n2v.py --datasetname Amazon-Photo --k 5

get adj_knn_n2v_5.npy in temp/Amazon-Photo/

train AEGCL

python train.py --device cuda:0 --dataset Amazon-Photo --param local:amazon_photo.json

Feel free to replace Amazon-Photo to other datasets.

Reference:

Wen-Zhi Li, Chang-Dong Wang, Jian-Huang Lai, Philip S. Yu. "Towards Effective and Robust Graph Contrastive Learning with Graph Autoencoding", TKDE2023.