torch_geometric.nn.aggr.Aggregation

class Aggregation

Bases: Module

用于实现自定义聚合的抽象基类。

聚合可以通过一个index向量来执行，该向量定义了从输入元素到它们在输出中位置的映射：

值得注意的是，index 不需要排序（对于大多数聚合操作符）：

# Feature matrix holding 10 elements with 64 features each:
x = torch.randn(10, 64)

# Assign each element to one of three sets:
index = torch.tensor([0, 0, 1, 0, 2, 0, 2, 1, 0, 2])

output = aggr(x, index)  #  Output shape: [3, 64]

或者，可以通过一个称为ptr的“压缩”索引向量来实现聚合。在这里，同一集合中的元素需要在输入中分组在一起，而ptr定义了它们的边界：

# Feature matrix holding 10 elements with 64 features each:
x = torch.randn(10, 64)

# Define the boundary indices for three sets:
ptr = torch.tensor([0, 4, 7, 10])

output = aggr(x, ptr=ptr)  #  Output shape: [3, 64]

请注意，至少需要定义index或ptr中的一个。

Shapes:

• 输入: 节点特征 \((*, |\mathcal{V}|, F_{in})\) 或边特征 \((*, |\mathcal{E}|, F_{in})\), 索引向量 \((|\mathcal{V}|)\) 或 \((|\mathcal{E}|)\),

• 输出: 图特征 \((*, |\mathcal{G}|, F_{out})\) 或 节点特征 \((*, |\mathcal{V}|, F_{out})\)

forward(x: Tensor, index: Optional[Tensor] = None, ptr: Optional[Tensor] = None, dim_size: Optional[int] = None, dim: int = -2, max_num_elements: Optional[int] = None) → Tensor

前向传播。

Parameters:

• x (torch.Tensor) – The source tensor.

• index (torch.Tensor, optional) – The indices of elements for applying the aggregation. One of index or ptr must be defined. (default: None)

• ptr (torch.Tensor, optional) – If given, computes the aggregation based on sorted inputs in CSR representation. One of index or ptr must be defined. (default: None)

• dim_size (int, optional) – The size of the output tensor at dimension dim after aggregation. (default: None)

• dim (int, optional) – The dimension in which to aggregate. (default: -2)

• max_num_elements (Optional[int], default: None) – (int, optional): The maximum number of elements within a single aggregation group. (default: None)

Return type:

Tensor

reset_parameters()

重置模块的所有可学习参数。