IGraphInferenceModel<T>
Interfaces ยท AiDotNet.Interfaces
Marker interface for models that support graph-aware inference: bind an adjacency matrix,
forward node features, get per-node predictions. Implemented by
NodeClassificationModel and any custom GNN
(GCN, GAT, GraphSAGE, etc.) that follows the same shape.
Known implementations: NodeClassificationModel<T>
How It Works
Introduced to replace the concrete-class gate in AiModelResultGraphExtensions โ
extension methods now gate on this interface so custom GNN subclasses can plug in without
deriving from NodeClassificationModel.
Methods
| Method | Summary |
|---|---|
Tensor<T> ForwardOnGraph(Tensor<T> nodeFeatures) |
Runs a forward pass through the network. |
void SetAdjacencyMatrix(Tensor<T> adjacencyMatrix) |
Binds the graph structure into every graph-convolutional layer prior to inference. |