This repo supplements our paper published in ICML-24. We explored employing a distribution of parametrized graphs for training a GNN in an Expectation Maximization (EM) framework. Through a probabilistic framework, we handle the uncertainty in graph structures stemming from various sources. Our approach enables the model to handle multiple graphs.
Download datasets (DBLP, ACM, IMDB) from this link and extract data.zip into data folder. (reference: GTN)
Alternatively, unzip data.zip into data folder.
The result of each dataset is the corresponding text files out_ACM.txt, out_DBLP.txt, out_IMDB.txt, and out_CORA.txt.
python main_batch_gcn.py --usedataset=IMDB --gamma=4 --alpha=0.2 --T=15 --Tprime=25
python main_batch_gcn.py --usedataset=DBLP --gamma=4 --alpha=0.2 --T=15 --Tprime=25 --patience=200 --lr=0.003
python main_batch_gcn.py --usedataset=ACM --gamma=2 --alpha=0.2 --T=15 --Tprime=25 --patience=150 --lr=0.005 --layer=2
python main_homogeneous2.py --n-layers=2 --n-epochs=300 --gpu=1 --label_n_per_class=20 --dropout=0.8 --dataset=cora --self-loop