GCN

class eyefeatures.deep.models.GCN(num_nodes, feature_dim, embedding_dim, layer_sizes, out_channels, use_embeddings=True)[source]

Bases: Module

A graph convolutional network (GCN) with optional learnable node embeddings.

Parameters:

  • num_nodes – (int) Number of nodes in the graph.

  • feature_dim – (int) Dimensionality of node features.

  • embedding_dim – (int) Dimensionality of the learnable node embeddings.

  • layer_sizes – (List[int]) List of hidden layer sizes for each GCN layer.

  • out_channels – (int) Number of output channels.

  • use_embeddings – (bool, optional) Whether to use learnable embeddings. Default is True.

Returns:

Output tensor after graph convolutions and global mean pooling.