DeepGCNLayer

torchmil.nn.gnns.DeepGCNLayer

Bases: Module

Implementation of a DeepGCN layer.

Adapts the implementation from torch_geometric.

__init__(conv=None, norm=None, act=None, block='plain', dropout=0.0)

Parameters:

• conv (Module, default: None ) –

  Convolutional layer.

• norm (Module, default: None ) –

  Normalization layer.

• act (Module, default: None ) –

  Activation layer.

• block (str, default: 'plain' ) –

  Skip connection type. Possible values: 'res', 'res+', 'dense', 'plain'.

• dropout (float, default: 0.0 ) –

  Dropout rate.

forward(x, adj)

Forward method.

Parameters:

• x (Tensor) –

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

• adj (Tensor) –

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

Returns:

• y ( Tensor ) –

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