drgst

Distribution Recovered Graph Self-Training framework (DR-GST)

• Paper link: https://arxiv.org/abs/2201.11349
• Author's code repo: https://github.com/TaurusTaurus-Rui/DR-GST. Note that the original code is implemented with Pytorch for the paper.

Dataset Statics

Dataset	# Nodes	# Edges	# Classes
Cora	2,708	10,556	7
Citeseer	3,327	9,228	6
Pubmed	19,717	88,651	3
Flickr	---	899576	7

Refer to Planetoid, Flickr

Results

# available dataset: "Cora", "Citeseer", "Pubmed", "CS", "Physics", "Computers", "Photo"
TL_BACKEND="torch" python dr_gst_trainer.py --dataset Cora --model GCN --labelrate 20 --drop_method dropout --droprate 0.3
TL_BACKEND="torch" python dr_gst_trainer.py --dataset CiteSeer --model GCN --labelrate 20 --drop_method dropout --droprate 0.3
TL_BACKEND="torch" python dr_gst_trainer.py --dataset PubMed --model GCN --labelrate 20 --drop_method dropout --droprate 0.3
TL_BACKEND="torch" python dr_gst_trainer.py --dataset CoraFull --model GCN --labelrate 20 --drop_method dropout --droprate 0.3
TL_BACKEND="torch" python dr_gst_trainer.py --model GCN --labelrate 5 --drop_method dropout --droprate 0.4 --dataset Flickr--threshold 0.8 --weight_decay 5e-5 --lr 0.005
‘’‘


| Dataset  | Paper Code | Out(th) |
|----------|------------|---------|
| Cora     | 83.34      | 88.60   |
| CiteSeer | 75.78      | 77.60   |
| PubMed   | 81.08      | 88.70   | 
| CoraFull | 62.75      | 80.30   |
| Flickr   | 37.84      | 37.59   |