from typing import Callable, Union
from torch_geometric.typing import OptPairTensor, Adj, OptTensor, Size
import torch
from torch import Tensor
import torch.nn.functional as F
from torch_sparse import SparseTensor, matmul
from torch_geometric.nn.conv import MessagePassing
from ..inits import reset
[docs]class GINConv(MessagePassing):
r"""The graph isomorphism operator from the `"How Powerful are
Graph Neural Networks?" <https://arxiv.org/abs/1810.00826>`_ paper
.. math::
\mathbf{x}^{\prime}_i = h_{\mathbf{\Theta}} \left( (1 + \epsilon) \cdot
\mathbf{x}_i + \sum_{j \in \mathcal{N}(i)} \mathbf{x}_j \right)
or
.. math::
\mathbf{X}^{\prime} = h_{\mathbf{\Theta}} \left( \left( \mathbf{A} +
(1 + \epsilon) \cdot \mathbf{I} \right) \cdot \mathbf{X} \right),
here :math:`h_{\mathbf{\Theta}}` denotes a neural network, *.i.e.* an MLP.
Args:
nn (torch.nn.Module): A neural network :math:`h_{\mathbf{\Theta}}` that
maps node features :obj:`x` of shape :obj:`[-1, in_channels]` to
shape :obj:`[-1, out_channels]`, *e.g.*, defined by
:class:`torch.nn.Sequential`.
eps (float, optional): (Initial) :math:`\epsilon`-value.
(default: :obj:`0.`)
train_eps (bool, optional): If set to :obj:`True`, :math:`\epsilon`
will be a trainable parameter. (default: :obj:`False`)
**kwargs (optional): Additional arguments of
:class:`torch_geometric.nn.conv.MessagePassing`.
"""
def __init__(self, nn: Callable, eps: float = 0., train_eps: bool = False,
**kwargs):
kwargs.setdefault('aggr', 'add')
super(GINConv, self).__init__(**kwargs)
self.nn = nn
self.initial_eps = eps
if train_eps:
self.eps = torch.nn.Parameter(torch.Tensor([eps]))
else:
self.register_buffer('eps', torch.Tensor([eps]))
self.reset_parameters()
[docs] def reset_parameters(self):
reset(self.nn)
self.eps.data.fill_(self.initial_eps)
[docs] def forward(self, x: Union[Tensor, OptPairTensor], edge_index: Adj,
size: Size = None) -> Tensor:
""""""
if isinstance(x, Tensor):
x: OptPairTensor = (x, x)
# propagate_type: (x: OptPairTensor)
out = self.propagate(edge_index, x=x, size=size)
x_r = x[1]
if x_r is not None:
out += (1 + self.eps) * x_r
return self.nn(out)
def message(self, x_j: Tensor) -> Tensor:
return x_j
def message_and_aggregate(self, adj_t: SparseTensor,
x: OptPairTensor) -> Tensor:
adj_t = adj_t.set_value(None, layout=None)
return matmul(adj_t, x[0], reduce=self.aggr)
def __repr__(self):
return '{}(nn={})'.format(self.__class__.__name__, self.nn)
[docs]class GINEConv(MessagePassing):
r"""The modified :class:`GINConv` operator from the `"Strategies for
Pre-training Graph Neural Networks" <https://arxiv.org/abs/1905.12265>`_
paper
.. math::
\mathbf{x}^{\prime}_i = h_{\mathbf{\Theta}} \left( (1 + \epsilon) \cdot
\mathbf{x}_i + \sum_{j \in \mathcal{N}(i)} \mathrm{ReLU}
( \mathbf{x}_j + \mathbf{e}_{j,i} ) \right)
that is able to incorporate edge features :math:`\mathbf{e}_{j,i}` into
the aggregation procedure.
Args:
nn (torch.nn.Module): A neural network :math:`h_{\mathbf{\Theta}}` that
maps node features :obj:`x` of shape :obj:`[-1, in_channels]` to
shape :obj:`[-1, out_channels]`, *e.g.*, defined by
:class:`torch.nn.Sequential`.
eps (float, optional): (Initial) :math:`\epsilon`-value.
(default: :obj:`0.`)
train_eps (bool, optional): If set to :obj:`True`, :math:`\epsilon`
will be a trainable parameter. (default: :obj:`False`)
**kwargs (optional): Additional arguments of
:class:`torch_geometric.nn.conv.MessagePassing`.
"""
def __init__(self, nn: Callable, eps: float = 0., train_eps: bool = False,
**kwargs):
kwargs.setdefault('aggr', 'add')
super(GINEConv, self).__init__(**kwargs)
self.nn = nn
self.initial_eps = eps
if train_eps:
self.eps = torch.nn.Parameter(torch.Tensor([eps]))
else:
self.register_buffer('eps', torch.Tensor([eps]))
self.reset_parameters()
[docs] def reset_parameters(self):
reset(self.nn)
self.eps.data.fill_(self.initial_eps)
[docs] def forward(self, x: Union[Tensor, OptPairTensor], edge_index: Adj,
edge_attr: OptTensor = None, size: Size = None) -> Tensor:
""""""
if isinstance(x, Tensor):
x: OptPairTensor = (x, x)
# Node and edge feature dimensionalites need to match.
if isinstance(edge_index, Tensor):
assert edge_attr is not None
assert x[0].size(-1) == edge_attr.size(-1)
elif isinstance(edge_index, SparseTensor):
assert x[0].size(-1) == edge_index.size(-1)
# propagate_type: (x: OptPairTensor, edge_attr: OptTensor)
out = self.propagate(edge_index, x=x, edge_attr=edge_attr, size=size)
x_r = x[1]
if x_r is not None:
out += (1 + self.eps) * x_r
return self.nn(out)
def message(self, x_j: Tensor, edge_attr: Tensor) -> Tensor:
return F.relu(x_j + edge_attr)
def __repr__(self):
return '{}(nn={})'.format(self.__class__.__name__, self.nn)