You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think it might be cool and useful to have the option to pass a float in (0, 1] for num_neighbors. If nodes have have highly varied degrees, a raw count of, say, 5 neighbors may be a substantial amount of information node u and significantly less information for node v. It may be more consistent to sample 0.5 of u's neighbors and 0.5 of v's neighbors.
Alternatives
I actually have a very hard time wrapping my brain around writing my own PyG loaders. Idk if that's because the code is scattered across files or what, but I'm not sure where I would begin to write my own.
Additional context
No response
The text was updated successfully, but these errors were encountered:
🚀 The feature, motivation and pitch
I think it might be cool and useful to have the option to pass a float in (0, 1] for num_neighbors. If nodes have have highly varied degrees, a raw count of, say, 5 neighbors may be a substantial amount of information node u and significantly less information for node v. It may be more consistent to sample 0.5 of u's neighbors and 0.5 of v's neighbors.
Alternatives
I actually have a very hard time wrapping my brain around writing my own PyG loaders. Idk if that's because the code is scattered across files or what, but I'm not sure where I would begin to write my own.
Additional context
No response
The text was updated successfully, but these errors were encountered: