GCNConv

torchmil.nn.gnns.GCNConv

Bases: Module

Implementation of a Graph Convolutional Network (GCN) layer.

Adapts the implementation from torch_geometric.

__init__(in_dim, out_dim=None, add_self_loops=False, learn_weights=False, layer_norm=False, normalize=False, dropout=0.0, activation=torch.nn.Identity(), bias=True)

Parameters:

• in_dim (int) –

  Input dimension.

• out_dim (int, default: None ) –

  Output dimension.

• add_self_loops (bool, default: False ) –

  Whether to add self-loops.

• learn_weights (bool, default: False ) –

  Whether to use a linear layer after the convolution.

• layer_norm (bool, default: False ) –

  Whether to use layer normalization.

• normalize (bool, default: False ) –

  Whether to l2-normalize the output.

• dropout (float, default: 0.0 ) –

  Dropout rate.

• activation (Module, default: Identity() ) –

  Activation function to apply after the convolution.

• bias (bool, default: True ) –

  Whether to use bias.

forward(x, adj)

Parameters:

• x –

  Node features of shape (batch_size, n_nodes, in_dim).

• adj –

  Adjacency matrix of shape (batch_size, n_nodes, n_nodes).

Returns:

• y ( Tensor ) –

  Output tensor of shape (batch_size, n_nodes, out_dim).