The PyTorch implementation of Self-supervised Contrastive Graph Representation with Node and Graph Augmentation.

In this work, we propose a new graph augmentation method to generate an augmentation graph without changing any structures from the original graph. Meanwhile, a node augmentation method is proposed to augment the positive node pairs by discovering the most similar nodes in the same graph.

• torch==1.10.1+cu113
• torch_geometric==2.0.2
• scikit_learn==1.0.2

Install all dependencies using

pip install -r requirements.txt

You can use the following command, and the parameters are given

python train.py --dataset DBLP

The --dataset argument should be one of [Cora, CiteSeer, PubMed, DBLP].

The code refers to the following two papers. Thank them very much for their open source work.

Deep Graph Contrastive Representation Learning(GRACE)

Directed Graph Contrastive Learning(DiGCL)