mindspore.nn.MarginRankingLoss

class mindspore.nn.MarginRankingLoss(margin=0.0, reduction='mean')

MarginRankingLoss creates a criterion that measures the loss.

Given two tensors \(input1\), \(input2\) and a Tensor label \(target\) with values 1 or -1, the operation is as follows:

\[\text{loss}(input1, input2, target) = \max(0, -target * (input1 - input2) + \text{margin})\]

Parameters

• margin (float, optional) – Specify the adjustment factor of the operation. Default: 0.0 .

• reduction (str, optional) –

  Apply specific reduction method to the output: 'none' , 'mean' , 'sum' . Default: 'mean' .

  • 'none': no reduction will be applied.

  • 'mean': compute and return the mean of elements in the output.

  • 'sum': the output elements will be summed.

Inputs:

• input1 (Tensor) - Tensor of shape \((N, *)\) where \(*\) means, any number of additional dimensions.

• input2 (Tensor) - Tensor of shape \((N, *)\), same shape and dtype as input1.

• target (Tensor) - Contains value 1 or -1. Suppose the shape of input1 is \((x_1, x_2, x_3, ..., x_R)\), then the shape of target must be \((x_1, x_2, x_3, ..., x_R)\).

Outputs:

Tensor or Scalar. if reduction is "none", its shape is the same as labels. Otherwise, a scalar value will be returned.

Raises

• TypeError – If margin is not a float.

• TypeError – If input1, input2 or target is not a Tensor.

• TypeError – If the types of input1 and input2 are inconsistent.

• TypeError – If the types of input1 and target are inconsistent.

• ValueError – If the shape of input1 and input2 are inconsistent.

• ValueError – If the shape of input1 and target are inconsistent.

• ValueError – If reduction is not one of 'none', 'mean' , 'sum'.

Supported Platforms:

Ascend GPU CPU

Examples

>>> import mindspore as ms
>>> from mindspore import Tensor, nn, ops
>>> import numpy as np
>>> loss1 = nn.MarginRankingLoss(reduction='none')
>>> loss2 = nn.MarginRankingLoss(reduction='mean')
>>> loss3 = nn.MarginRankingLoss(reduction='sum')
>>> sign = ops.Sign()
>>> input1 = Tensor(np.array([0.3864, -2.4093, -1.4076]), ms.float32)
>>> input2 = Tensor(np.array([-0.6012, -1.6681, 1.2928]), ms.float32)
>>> target = sign(Tensor(np.array([-2, -2, 3]), ms.float32))
>>> output1 = loss1(input1, input2, target)
>>> print(output1)
[0.98759997 0.         2.7003999 ]
>>> output2 = loss2(input1, input2, target)
>>> print(output2)
1.2293333
>>> output3 = loss3(input1, input2, target)
>>> print(output3)
3.6879997