PyTorch implementation of FastGCN and AS-GCN. The supported datasets are: cora, citeseer and pubmed. Mind that this implementation may differ from the original in some parts. Especially for the AS-GCN, the different methods of calculating variance did not bring better performance. So if you want to use it into the research, please cheak these details carefully.
* PyTorch 1.14
* Python 3.7
python train.py --dataset dataset_name --model model_name
FASTGCN: FAST LEARNING WITH GRAPH CONVOLUTIONAL NETWORKS VIA IMPORTANCE SAMPLING
Adaptive Sampling Towards Fast Graph Representation Learning