GraphPooling
Enums · AiDotNet.Enums
Defines how a graph model combines its per-node representations into a single representation for the whole graph.
For Beginners
A graph model produces one summary per node, but labelling the whole graph needs a single summary. Pooling is the step that combines them. Averaging is the usual choice and treats every node equally; the others emphasise the strongest signal, the total amount of signal, or let the model learn which nodes matter.
How It Works
Promoted out of GraphClassificationModel<T>, where it was nested. A type nested
in a generic class is a distinct type per type argument, so
GraphClassificationModel<double>.GraphPooling and the <float> one
were unrelated types and the value could not be named on a non-generic options class.
Fields
| Field | Summary |
|---|---|
GraphPooling Attention |
Weighted average with learned attention over the nodes. |
GraphPooling Max |
Take the maximum across all node embeddings. |
GraphPooling Mean |
Average all node embeddings. |
GraphPooling Sum |
Sum all node embeddings, so larger graphs produce larger values. |
int value__ |