merit

GammaGL Implementation of MERIT

This GammaGL example implements the model proposed in the paper Multi-Scale Contrastive Siamese Networks for Self-Supervised Graph Representation Learning.

Author's code: https://github.com/GRAND-Lab/MERIT

Example Implementor

This example was implemented by Ziyu Zheng

Datasets

Unsupervised Node Classification Datasets:

'Cora', 'Citeseer' and 'Pubmed'

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

Arguments

--input_dim 			int		Input dimension.                       Default is 1433.
--out_dim				int		Output dimension.					   Default is 512.
--proj_size 			int		Encoder output dimension			   Default is 512.
--proj_hid 				int     Encoder hidden dimension			   Default is 4096.
--pred_size 			int		MLP output dimension			       Default is 512.
--pred_hid 				int		MLP hidden dimension			   	   Default is 4096.
--drop_edge_rate_1      float   Drop edge ratio 1.                     Default is 0.2. 
--drop_edge_rate_2      float   Drop edge ratio 2.                     Default is 0.2. 
--drop_feature_rate_1   float   Drop feature ratio 1.                  Default is 0.5. 
--drop_feature_rate_2   float   Drop feature ratio 2.                  Default is 0.5. 
--dataset_path          str     path to save dataset.                  Default is r'../'

How to run examples

In the paper(as well as authors' repo), the training set are full graph training

# use paddle backend
# Cora by GammaGL
TL_BACKEND=paddle python merit_trainer.py --dataset cora --epochs 500 --drop_edge_rate_1 0.2 --drop_edge_rate_2 0.2 --drop_feature_rate_1 0.5 --drop_feature_rate_2 0.5 --lr 3e-4 --beta 0.5
#Citeseer by GammaGL
TL_BACKEND=paddle python merit_trainer.py --dataset citeseer --epochs 500 --drop_edge_rate_1 0.4 --drop_edge_rate_2 0.4 --drop_feature_rate_1 0.5 --drop_feature_rate_2 0.5 --lr 3e-4 --beta 0.6

# use tensorflow backend
# Cora by GammaGL
TL_BACKEND=tensorflow python merit_trainer.py --dataset cora --epochs 500 --drop_edge_rate_1 0.2 --drop_edge_rate_2 0.2 --drop_feature_rate_1 0.5 --drop_feature_rate_2 0.5 --lr 3e-4 --beta 0.5
#Citeseer by GammaGL
TL_BACKEND=tensorflow python merit_trainer.py --dataset citeseer --epochs 500 --drop_edge_rate_1 0.4 --drop_edge_rate_2 0.4 --drop_feature_rate_1 0.5 --drop_feature_rate_2 0.5 --lr 3e-4 --beta 0.6

Performance

Dataset	Cora	Citeseer	Pubmed
Author's Code	83.1	74.0	80.2
GammaGL(tf)	84.3	72.2	--.-
GammaGL(paddle)	83.1	--.-	--.-