GCN

Model variant implementing a graph convolutional network for graph-level prediction. Each layer performs a graph convolution that mixes node features along edges, followed by a linear projection and activation; the final node features are mean-pooled per graph and projected to the output width.

Note

GCN uses scalar edge weights, so it expects edge attributes of shape [E, 1].

Configuration

Selected as name: gcn.

Option	Description	Type	Default
hidden_layers	Number of graph-convolution blocks.	int	required
hidden_channels	Width of each hidden layer.	int	required

components:
  model:
    name: gcn
    kwargs:
      hidden_layers: 4
      hidden_channels: 256