torch_geometric.nn.conv.FiLMConv
- class FiLMConv(in_channels: Union[int, Tuple[int, int]], out_channels: int, num_relations: int = 1, nn: Optional[Callable] = None, act: Optional[Callable] = ReLU(), aggr: str = 'mean', **kwargs)[source]
Bases:
MessagePassing来自“GNN-FiLM: 具有特征线性调制的图神经网络”论文的FiLM图卷积操作符。
\[\mathbf{x}^{\prime}_i = \sum_{r \in \mathcal{R}} \sum_{j \in \mathcal{N}(i)} \sigma \left( \boldsymbol{\gamma}_{r,i} \odot \mathbf{W}_r \mathbf{x}_j + \boldsymbol{\beta}_{r,i} \right)\]其中 \(\boldsymbol{\beta}_{r,i}, \boldsymbol{\gamma}_{r,i} = g(\mathbf{x}_i)\)，默认情况下 \(g\) 是一个单层线性层。自环会自动添加到输入图中，并表示为它自己的关系类型。
注意
有关使用FiLM的示例，请参见examples/gcn.py。
- Parameters:
in_channels (int or tuple) – Size of each input sample, or
-1to 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.
num_relations (int, optional) – 关系的数量。（默认值：
1）nn (torch.nn.Module, optional) – 将节点特征
x_i从形状[-1, in_channels]映射到形状[-1, 2 * out_channels]的神经网络 \(g\)。如果设置为None，\(g\) 将被实现为单个线性层。（默认值：None）act (可调用的, 可选的) – 激活函数 \(\sigma\). (默认:
torch.nn.ReLU())aggr (str, optional) – The aggregation scheme to use (
"add","mean","max"). (default:"mean")**kwargs (optional) – Additional arguments of
torch_geometric.nn.conv.MessagePassing.
- Shapes:
输入: 节点特征 \((|\mathcal{V}|, F_{in})\) 或 \(((|\mathcal{V_s}|, F_{s}), (|\mathcal{V_t}|, F_{t}))\) 如果是二分图, 边索引 \((2, |\mathcal{E}|)\), 边类型 \((|\mathcal{E}|)\)
输出: 节点特征 \((|\mathcal{V}|, F_{out})\) 或 \((|\mathcal{V_t}|, F_{out})\) 如果是二分图