Adversarial Contrastive Graph Augmentation with Counterfactual Regularization
December 17, 2024 · View on GitHub
Implementation for AAAI'25 paper: Adversarial Contrastive Graph Augmentation with Counterfactual Regularization
environment install:
pip install -r requirements.txt
Parameter introduction:
--dataset cora or citeseer or airpotr ...
--hidden_size encoder hidden dimension, node clsssification defalut 128,link prediction defalut 32
--emb_size VGAE/GAE embedding dimension,node clsssification defalut 32,link prediction defalut 16
--gae selct VGAE or GAE
--use_bns use bn
--task 0 is node clsssification, and 1 is link prediction
--alpha
--beta
--gamma optuna get best parameter
Run parameter search node classification:
python optuna_ACGA.py --dataset cora --hidden_size 128 --emb_size 32 --gae 0 --use_bns True --task 0
Run parameter search link prediction:
python optuna_ACGA.py --dataset cora --hidden_size 32 --emb_size 16 --gae 0 --use_bns True --task 1
Run the node classification task:
python ACGA.py --dataset cora --hidden_size 128 --emb_size 32 --gae 0 --use_bns True --task 0
Run the link prediction task:
python ACGA.py --dataset cora --hidden_size 32 --emb_size 16 --gae 0 --use_bns True --task 1