IF YOU WOULD LIKE TO GET AN ACCOUNT, please write an email to s dot adaszewski at gmail dot com. User accounts are meant only to report issues and/or generate pull requests. This is a purpose-specific Git hosting for ADARED projects. Thank you for your understanding!
Du kan inte välja fler än 25 ämnen Ämnen måste starta med en bokstav eller siffra, kan innehålla bindestreck ('-') och vara max 35 tecken långa.

54 lines
1.8KB

  1. import torch
  2. from .dropout import dropout_sparse
  3. from .weights import init_glorot
  4. class SparseGraphConv(torch.nn.Module):
  5. """Convolution layer for sparse inputs."""
  6. def __init__(self, in_channels, out_channels,
  7. adjacency_matrix, **kwargs):
  8. super().__init__(**kwargs)
  9. self.in_channels = in_channels
  10. self.out_channels = out_channels
  11. self.weight = init_glorot(in_channels, out_channels)
  12. self.adjacency_matrix = adjacency_matrix
  13. def forward(self, x):
  14. x = torch.sparse.mm(x, self.weight)
  15. x = torch.sparse.mm(self.adjacency_matrix, x)
  16. return x
  17. class SparseDropoutGraphConvActivation(torch.nn.Module):
  18. def __init__(self, input_dim, output_dim,
  19. adjacency_matrix, keep_prob=1.,
  20. activation=torch.nn.functional.relu,
  21. **kwargs):
  22. super().__init__(**kwargs)
  23. self.sparse_graph_conv = SparseGraphConv(input_dim, output_dim)
  24. self.keep_prob = keep_prob
  25. self.activation = activation
  26. def forward(self, x):
  27. x = dropout_sparse(x, self.keep_prob)
  28. x = self.sparse_graph_conv(x)
  29. x = self.activation(x)
  30. return x
  31. class SparseMultiDGCA(torch.nn.Module):
  32. def __init__(self, input_dim, output_dim,
  33. adjacency_matrices, keep_prob=1.,
  34. activation=torch.nn.functional.relu,
  35. **kwargs):
  36. super().__init__(**kwargs)
  37. self.sparse_dgca = [ SparseDropoutGraphConvActivation(input_dim, output_dim, adj_mat, keep_prob, activation) for adj_mat in adjacency_matrices ]
  38. def forward(self, x):
  39. out = torch.zeros(len(x), output_dim, dtype=x.dtype)
  40. for f in self.sparse_dgca:
  41. out += f(x)
  42. out = torch.nn.functional.normalize(out, p=2, dim=1)
  43. return out