Hi. First of all, I’m a beginner at GNN, and my English might be weird. I would appreciate your understanding. My data set consists of a node for the user and an edge for the transaction. When we created the data, we made a two-step graph around one node. I’m going to use the EGAT layer to classify graphs, but there are cases where one user appears in each graph in duplicate. In this case, does message passing overlap? How can I handle it? Thank you.
Did you mean that some nodes can appear in multiple graphs simultaneously? In such cases, it might help to merge all graphs into one single graph for message passing and then perform graph classification with the updated node representations. However, since you are a beginner, it might be easier to forget about it and treat the nodes/graphs as independent objects.
Thanks for you reply, mufeili.
I want to merge the graph but my dataset has over than 5 million nodes and 10 million edges. I think it is too big number to merge all graph. I will try the 2nd recommend (independently).
1 Like
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.