mindspore.nn.MultilabelMarginLoss

class mindspore.nn.MultilabelMarginLoss(reduction='mean')[source]

Creates a loss criterion that minimizes the hinge loss for multi-class classification tasks. It takes a 2D mini-batch Tensor \(x\) as input and a 2D Tensor \(y\) containing target class indices as output.

Each sample in the mini-batch, the loss is computed as follows:

\[\text{loss}(x, y) = \sum_{ij}\frac{\max(0, 1 - (x[y[j]] - x[i]))}{\text{x.size}(0)}\]

where \(x \in \left\{0, \; \cdots , \; \text{x.size}(0) - 1\right\}\), \(y \in \left\{0, \; \cdots , \; \text{y.size}(0) - 1\right\}\), \(0 \leq y[j] \leq \text{x.size}(0)-1\), and for all \(i\) and \(j\), \(i\) does not equal to \(y[j]\).

Furthermore, both \(y\) and \(x\) should have identical sizes.

Note

For this operator, only a contiguous sequence of non-negative targets that starts at the beginning is taken into consideration, which means that different samples can have different number of target classes.

Parameters

reduction (str, optional) – Apply specific reduction method to the output: 'none' , 'mean' , 'sum' . Default: "mean" .

Inputs:

  • x (Tensor) - Predict data. Tensor of shape \((C)\) or \((N, C)\), where \(N\) is the batch size and \(C\) is the number of classes. Data type must be float16 or float32.

  • target (Tensor) - Ground truth data, with the same shape as x, data type must be int32 and label targets padded by -1.

Outputs:

  • y (Union[Tensor, Scalar]) - The loss of MultilabelMarginLoss. If reduction is “none”, its shape is \((N)\). Otherwise, a scalar value will be returned.

Raises

  • TypeError – If x or target is not a Tensor.

  • TypeError – If dtype of x is neither float16 nor float32.

  • TypeError – If dtype of target is not int32.

  • ValueError – If length of shape of x is neither 1 nor 2.

  • ValueError – If shape of x is not the same as target.

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

Supported Platforms:

Ascend GPU

Examples

>>> import mindspore as ms
>>> import mindspore.nn as nn
>>> import numpy as np
>>> loss = nn.MultilabelMarginLoss()
>>> x = ms.Tensor(np.array([[0.1, 0.2, 0.4, 0.8], [0.2, 0.3, 0.5, 0.7]]), ms.float32)
>>> target = ms.Tensor(np.array([[1, 2, 0, 3], [2, 3, -1, 1]]), ms.int32)
>>> output = loss(x, target)
>>> print(output)
0.325