半径图

class dgl.nn.pytorch.factory.RadiusGraph(r, p=2, self_loop=False, compute_mode='donot_use_mm_for_euclid_dist')

Bases: Module

将一组点转换为在给定距离内具有邻居的双向图的层。

RadiusGraph 的实现步骤如下：

1. 计算所有点的NxN成对距离矩阵。

2. 选择距离每个点在一定范围内的点作为它们的邻居。

3. 构建一个图，其中每个点作为节点与其邻居相连。

返回的图的节点对应于点，其中每个点的邻居在给定距离内。

Parameters:

• r (float) – Radius of the neighbors.

• p (float, optional) –

  Power parameter for the Minkowski metric. When p = 1 it is the equivalent of Manhattan distance (L1 norm) and Euclidean distance (L2 norm) for p = 2.

  (default: 2)

• self_loop (bool, optional) –

  Whether the radius graph will contain self-loops.

  (default: False)

• compute_mode (str, optional) –

  use_mm_for_euclid_dist_if_necessary - will use matrix multiplication approach to calculate euclidean distance (p = 2) if P > 25 or R > 25 use_mm_for_euclid_dist - will always use matrix multiplication approach to calculate euclidean distance (p = 2) donot_use_mm_for_euclid_dist - will never use matrix multiplication approach to calculate euclidean distance (p = 2).

  (default: donot_use_mm_for_euclid_dist)

示例

以下示例使用 PyTorch 后端。

>>> import dgl
>>> from dgl.nn.pytorch.factory import RadiusGraph

>>> x = torch.tensor([[0.0, 0.0, 1.0],
...                   [1.0, 0.5, 0.5],
...                   [0.5, 0.2, 0.2],
...                   [0.3, 0.2, 0.4]])
>>> rg = RadiusGraph(0.75)
>>> g = rg(x)  # Each node has neighbors within 0.75 distance
>>> g.edges()
(tensor([0, 1, 2, 2, 3, 3]), tensor([3, 2, 1, 3, 0, 2]))

当 get_distances 为 True 时，前向传递返回半径图和对应边的距离。

>>> x = torch.tensor([[0.0, 0.0, 1.0],
...                   [1.0, 0.5, 0.5],
...                   [0.5, 0.2, 0.2],
...                   [0.3, 0.2, 0.4]])
>>> rg = RadiusGraph(0.75)
>>> g, dist = rg(x, get_distances=True)
>>> g.edges()
(tensor([0, 1, 2, 2, 3, 3]), tensor([3, 2, 1, 3, 0, 2]))
>>> dist
tensor([[0.7000],
        [0.6557],
        [0.6557],
        [0.2828],
        [0.7000],
        [0.2828]])

forward(x, get_distances=False)

前向计算。

Parameters:

• x (Tensor) – 点的坐标。 \((N, D)\) 其中 \(N\) 表示 点集中点的数量，\(D\) 表示 特征的大小。它可以在CPU或GPU上。点坐标的设备 指定了半径图的设备。

• get_distances (bool, optional) –

  Whether to return the distances for the corresponding edges in the radius graph.

  (default: False)

Returns:

• DGLGraph – 构建的图。节点ID的顺序与x相同。

• torch.Tensor, optional – 构建图中边的距离。距离的顺序与边ID的顺序相同。