GIN

class eyefeatures.deep.models.GIN(num_common_nodes, feature_dim, embedding_dim, layer_sizes, out_channels)

Bases: Module

A graph isomorphism network (GIN) with shared embeddings for nodes across graphs.

Parameters:

• num_common_nodes – (int) Number of shared nodes across graphs.

• feature_dim – (int) Dimensionality of node features.

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

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

• out_channels – (int) Number of output channels.

Returns:

Output tensor after GIN layers and global mean pooling.