Dear authors,
thank you for open-sourcing your code :) I am planning to test your approach on temporal graphs that differ wildly in terms graph size, i.e., number of nodes and edges. Do you generally think that TGN's are very sensitive to varying graph sizes (especially at initialization) or did that not come up when developing your model?
Best regards
Dear authors,
thank you for open-sourcing your code :) I am planning to test your approach on temporal graphs that differ wildly in terms graph size, i.e., number of nodes and edges. Do you generally think that TGN's are very sensitive to varying graph sizes (especially at initialization) or did that not come up when developing your model?
Best regards