from typing import Callable, Optional, Tuple, Union
from torch import Tensor
from torch.nn import Parameter, Sigmoid
from torch_geometric.nn.conv import MessagePassing
from torch_geometric.nn.dense.linear import Linear
from torch_geometric.typing import Adj, PairTensor
from ..inits import zeros
[docs]class ResGatedGraphConv(MessagePassing):
r"""The residual gated graph convolutional operator from the
`"Residual Gated Graph ConvNets" <https://arxiv.org/abs/1711.07553>`_ paper
.. math::
\mathbf{x}^{\prime}_i = \mathbf{W}_1 \mathbf{x}_i +
\sum_{j \in \mathcal{N}(i)} \eta_{i,j} \odot \mathbf{W}_2 \mathbf{x}_j
where the gate :math:`\eta_{i,j}` is defined as
.. math::
\eta_{i,j} = \sigma(\mathbf{W}_3 \mathbf{x}_i + \mathbf{W}_4
\mathbf{x}_j)
with :math:`\sigma` denoting the sigmoid function.
Args:
in_channels (int or tuple): Size of each input sample, or :obj:`-1` to
derive the size from the first input(s) to the forward method.
A tuple corresponds to the sizes of source and target
dimensionalities.
out_channels (int): Size of each output sample.
act (callable, optional): Gating function :math:`\sigma`.
(default: :meth:`torch.nn.Sigmoid()`)
bias (bool, optional): If set to :obj:`False`, the layer will not learn
an additive bias. (default: :obj:`True`)
root_weight (bool, optional): If set to :obj:`False`, the layer will
not add transformed root node features to the output.
(default: :obj:`True`)
**kwargs (optional): Additional arguments of
:class:`torch_geometric.nn.conv.MessagePassing`.
Shapes:
- **inputs:**
node features :math:`(|\mathcal{V}|, F_{in})` or
:math:`((|\mathcal{V_s}|, F_{s}), (|\mathcal{V_t}|, F_{t}))`
if bipartite,
edge indices :math:`(2, |\mathcal{E}|)`
- **outputs:** node features :math:`(|\mathcal{V}|, F_{out})` or
:math:`(|\mathcal{V_t}|, F_{out})` if bipartite
"""
def __init__(
self,
in_channels: Union[int, Tuple[int, int]],
out_channels: int,
act: Optional[Callable] = Sigmoid(),
root_weight: bool = True,
bias: bool = True,
**kwargs,
):
kwargs.setdefault('aggr', 'add')
super().__init__(**kwargs)
self.in_channels = in_channels
self.out_channels = out_channels
self.act = act
self.root_weight = root_weight
if isinstance(in_channels, int):
in_channels = (in_channels, in_channels)
self.lin_key = Linear(in_channels[1], out_channels)
self.lin_query = Linear(in_channels[0], out_channels)
self.lin_value = Linear(in_channels[0], out_channels)
if root_weight:
self.lin_skip = Linear(in_channels[1], out_channels, bias=False)
else:
self.register_parameter('lin_skip', None)
if bias:
self.bias = Parameter(Tensor(out_channels))
else:
self.register_parameter('bias', None)
self.reset_parameters()
[docs] def reset_parameters(self):
self.lin_key.reset_parameters()
self.lin_query.reset_parameters()
self.lin_value.reset_parameters()
if self.lin_skip is not None:
self.lin_skip.reset_parameters()
if self.bias is not None:
zeros(self.bias)
[docs] def forward(self, x: Union[Tensor, PairTensor], edge_index: Adj) -> Tensor:
""""""
if isinstance(x, Tensor):
x: PairTensor = (x, x)
k = self.lin_key(x[1])
q = self.lin_query(x[0])
v = self.lin_value(x[0])
# propagate_type: (k: Tensor, q: Tensor, v: Tensor)
out = self.propagate(edge_index, k=k, q=q, v=v, size=None)
if self.root_weight:
out = out + self.lin_skip(x[1])
if self.bias is not None:
out = out + self.bias
return out
def message(self, k_i: Tensor, q_j: Tensor, v_j: Tensor) -> Tensor:
return self.act(k_i + q_j) * v_j