scglue.models.nn.GraphConv.forward

GraphConv.forward(input, eidx, enorm, esgn)[source]

Forward propagation

Parameters

  • input (Tensor) – Input data (\(n_{vertices} \times n_{features}\))

  • eidx (Tensor) – Vertex indices of edges (\(2 \times n_{edges}\))

  • enorm (Tensor) – Normalized weight of edges (\(n_{edges}\))

  • esgn (Tensor) – Sign of edges (\(n_{edges}\))

Return type

Tensor

Returns

result – Graph convolution result (\(n_{vertices} \times n_{features}\))