mindspore.nn.GaussianNLLLoss

class mindspore.nn.GaussianNLLLoss(*, full=False, eps=1e-06, reduction='mean')[source]

Gaussian negative log likelihood loss.

The targets are treated as samples from Gaussian distributions with expectations and variances predicted by the neural network. For a labels tensor modelled as having Gaussian distribution with a tensor of expectations logits and a tensor of positive variances var the loss is:

\[\text{loss} = \frac{1}{2}\left(\log\left(\text{max}\left(\text{var}, \ \text{eps}\right)\right) + \frac{\left(\text{logits} - \text{labels}\right)^2} {\text{max}\left(\text{var}, \ \text{eps}\right)}\right) + \text{const.}\]

where eps is used for stability of \(log\). By default, the constant term of the loss function is omitted unless \(full=True\). If the shape of \(var\) is not the same as logits (due to a homoscedastic assumption), it must either have a final dimension of 1 or have one fewer dimension (with all other sizes being the same) for correct broadcasting.

Parameters

  • full (bool) – Include the constant term in the loss calculation. When \(full=True\), the constant term const. will be \(0.5 * log(2\pi)\). Default: False.

  • eps (float) – Used to improve the stability of log function. Default: 1e-6.

  • reduction (str) – Apply specific reduction method to the output: ‘none’, ‘mean’, or ‘sum’. Default: ‘mean’.

Inputs:

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

  • labels (Tensor) - Tensor of shape \((N, *)\) or \((*)\), same shape as the logits, or same shape as the logits but with one dimension equal to 1 (to allow for broadcasting).

  • var - Tensor of shape \((N, *)\) or \((*)\), same shape as logits, or same shape as the logits but with one dimension equal to 1, or same shape as the logits but with one fewer dimension (to allow for broadcasting).

Returns

Tensor or Tensor scalar, the computed loss depending on reduction.

Raises

  • TypeError – If logits is not a Tensor.

  • TypeError – If labels is not a Tensor.

  • TypeError – If full is not a bool.

  • TypeError – If eps is not a float.

  • ValueError – If eps is not a float within [0, inf).

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

Supported Platforms:

Ascend GPU CPU

Examples

>>> import numpy as np
>>> from mindspore import Tensor
>>> import mindspore.nn as nn
>>> import mindspore.common.dtype as mstype
>>> arr1 = np.arange(8).reshape((4, 2))
>>> arr2 = np.array([2, 3, 1, 4, 6, 4, 4, 9]).reshape((4, 2))
>>> logits = Tensor(arr1, mstype.float32)
>>> labels = Tensor(arr2, mstype.float32)
>>> loss = nn.GaussianNLLLoss(reduction='mean')
>>> var = Tensor(np.ones((4, 1)), mstype.float32)
>>> output = loss(logits, labels, var)
>>> print(output)
1.4374993
Reference:

Nix, D. A. and Weigend, A. S., “Estimating the mean and variance of the target probability distribution”, Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN’94), Orlando, FL, USA, 1994, pp. 55-60 vol.1, doi: 10.1109/ICNN.1994.374138.