How many K-hop neighborhood dgl consider if apply the sageconv layer from the DGL library?
Is it necessary to update the weights in GSAGE layer if we have edge weights along with edge index? Or, by default if we put the weights as input it automatically do calculation?
The number of neighbors during message passing is decided by your input graph (all direct neighbor nodes for the update of a target node).
The forward method of sageconv has an optional argument edge_weight. You can pass your edge weights to it.
thanks. if our graph has two hops or three hops neighbor, these neighbors doesn’t consider? Is there any way to consider two or three hops neighbors in DGL?
By stacking K conv layers for a model, the update of a specific target node will consider K-hop neighbors.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.