Replies: 1 comment
-
Inner product is certainly the most common since it allows you to do Maximum Inner Product Search (MIPS) on node embeddings. You can also use other k-NN tools, e.g., based on Euclidean distance). Alternatively, you can have a parametrized decoder, e.g., an MLP on source and destination node embeddings that maps to a link probability, but this is expensive to compute for O(N^2) pairs. |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Dear pyg users and developers,
Is there a way other than inner product to reconstruct the node-to-edge representation here?
If so, please let me know the paper or the exact method.
Beta Was this translation helpful? Give feedback.
All reactions