The overview of GCMAE.

Generative and Contrastive Paradigms Are Complementary for Graph Self-Supervised Learning

GCMAE is a self-supervised graph representation method, which unfies the contrastive learning and graph masked autoencoder. We conducted extensive experiments on various graph tasks, including node classification, link prediction, node clustering, and graph classification.

Dependencies

• Python >= 3.7
• Pytorch >= 1.9.0
• dgl >= 0.7.2
• pyyaml == 5.4.1
• munkres

Quick Start

For quick start, you could run the scripts:

Node classification

# Run the code manually for node classification:
python main.py --dataset cora --device 0

Link prediction

# Run the code manually for link prediction:
python main_lp.py --dataset cora --device 0 

Node clustering

# Run the code manually for node clustering:
python main.py --dataset cora --task cls --device 0 

Graph classification

# Run the code manually for graph classification:
python main_graph.py --dataset IMDB-BINARY --device 0 

Run with --use_cfg in command to reproduce the reported results.