Question about the subgraph generation module

[Hyewonon]

Hi, thanks for your job.

I have a question about the subgraph generation module.

According to the paper, the input in the subgraph generation module is e_u(0) + e_u(1), the embedding of user ID and the feature contained by aggregating local neighbor in the graph. But, the code shows that both user and item embeddings might be used as input. Please reply why both user and item are used when grouping users in the code.