Some questions about graphsage

[scottshufe]

Hi, I would like to ask some questions about graphsage (or SAGEConv):

1. Is NeighborSampler necessary for a graphsage model?
2. Could you please provide me with some graphsage examples for the graph classification task?

(By the way, it often takes me a lot of time to seek the example codes I want. It would be great if your team could provide an index for the example codes in README file in the example directory)

Thanks in advance :)

[rusty1s]

Thanks for your interest.

1. No, we disentangle the model (SAGEConv) from the scalability technique (NeighborSampler). That way, you can use SAEGConv for full-batch training, and can use any other PyG GNN model for neighbor sampling as well.
2. Here is an example GraphSAGE model for graph classification.

[scottshufe]

Wow, thank you so much!

[rusty1s]

(By the way, it often takes me a lot of time to seek the example codes I want. It would be great if your team could provide an index for the example codes in README file in the example directory)

I totally understand that. This is a great idea. Let me work on that.

[scottshufe]

😉😉