infograph

InfoGraph

• Paper link: https://arxiv.org/pdf/1908.01000
• Author's code repo (in Pytorch): https://github.com/fanyun-sun/InfoGraph

How to run

Run with following:

# use tensorflow bakcend
TL_BACKEND="tensorflow" python infograph_trainer.py --dataset MUTAG 
TL_BACKEND="tensorflow" python infograph_trainer.py --dataset IMDB-BINARY
TL_BACKEND="tensorflow" python infograph_trainer.py --dataset REDDIT-BINARY

# use pytorch backend
TL_BACKEND="torch" python infograph_trainer.py --dataset MUTAG 
TL_BACKEND="torch" python infograph_trainer.py --dataset IMDB-BINARY
TL_BACKEND="torch" python infograph_trainer.py --dataset REDDIT-BINARY 

# use paddle backend
TL_BACKEND="paddle" python infograph_trainer.py --dataset MUTAG 
TL_BACKEND="paddle" python infograph_trainer.py --dataset IMDB-BINARY
TL_BACKEND="paddle" python infograph_trainer.py --dataset REDDIT-BINARY

Results (accuracy)

Dataset	MUTAG	IMDB-B	REDDIT-B
Author's Code	89.0	73.0	82.5
DGL	89.88	72.7	88.5
GammaGL(tf)	89.2 ± 0.25	72.0	82.4
GammaGL(th)	90.65 ± 0.2	--.-	--.-
GammaGL(pd)	89.2 ± 0.53	72.13 ± 0.12	84.8
GammaGL(ms)	--.-	--.-	--.-