First, thanks to DGL, a great tool.
Graph Classification Tutorial, if I understand it correctly, is transductive learning. So why use the trained model to classify (inductive way) graph you haven’t seen before?
Transductive learning refers to the case where the features and connectivities for test nodes are also visible in the training process and this makes graph classification problems naturally inductive problems.
Thank you for the timely reply, which helped me correctly understand the meaning of inductive learning here.