diffpool
diffpool copied to clipboard
Test dataset
Hi, thanks for your work. The code is really hard for me now, could you give me a overall guide for two questions?
- When a graph in test dataset has a size that does not show in train dataset, the parameters got from the train set don't fit the "unseen" graph, then how the codes work? And I can't even find the test dataset in the codes, just validation dataset, why this?
- The main way to construct the graph dataset is using networkx module, are there any advantages here? Cause the GCN, GAT,ST-GCN don't use this, they seem to be more brief. Thanks.