scglue.models.nn.GraphAttent.forward

GraphAttent.forward(input, eidx, ewt, esgn)[source]

Forward propagation

Parameters:

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

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

  • ewt (torch.Tensor) – Weight of edges (\(n_{edges}\))

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

Return type:

torch.Tensor

Returns:

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