PairwiseDistance¶

class paddle.nn. PairwiseDistance ( p=2.0, epsilon=1e-06, keepdim=False, name=None ) [source]

It computes the pairwise distance between two vectors. The distance is calculated by p-oreder norm:

$$\Vert x \Vert _p = \left( \sum_{i=1}^n \vert x_i \vert ^ p \right) ^ {1/p}.$$

Parameters

• p (float, optional) – The order of norm. Default: $2.0$.

• epsilon (float, optional) – Add small value to avoid division by zero. Default: $1e-6$.

• keepdim (bool, optional) – Whether to reserve the reduced dimension in the output Tensor. The result tensor is one dimension less than the result of |x-y| unless keepdim is True. Default: False.

• name (str, optional) – For details, please refer to Name. Generally, no setting is required. Default: None.

Shape:

• x: $[N, D]$ or $[D]$, where $N$ is batch size, $D$ is the dimension of the data. Available data type is float16, float32, float64.

• y: $[N, D]$ or $[D]$, y have the same dtype as x.

• output: The same dtype as input tensor.

  • If keepdim is True, the output shape is $[N, 1]$ or $[1]$, depending on whether the input has data shaped as $[N, D]$.

  • If keepdim is False, the output shape is $[N]$ or $[]$, depending on whether the input has data shaped as $[N, D]$.

Examples

import paddle
x = paddle.to_tensor([[1., 3.], [3., 5.]], dtype=paddle.float64)
y = paddle.to_tensor([[5., 6.], [7., 8.]], dtype=paddle.float64)
dist = paddle.nn.PairwiseDistance()
distance = dist(x, y)
print(distance)
# Tensor(shape=[2], dtype=float64, place=Place(gpu:0), stop_gradient=True,
#        [4.99999860, 4.99999860])

forward ( x, y )

forward¶

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

• *inputs (tuple) – unpacked tuple arguments

• **kwargs (dict) – unpacked dict arguments

extra_repr ( )

extra_repr¶

Extra representation of this layer, you can have custom implementation of your own layer.