mindspore.nn.probability.distribution.Gumbel

class mindspore.nn.probability.distribution.Gumbel(loc, scale, seed=0, dtype=mstype.float32, name='Gumbel')[source]

Gumbel distribution. A Gumbel distributio is a continuous distribution with the range \([0, 1]\) and the probability density function:

\[f(x, a, b) = 1 / b \exp(\exp(-(x - a) / b) - x),\]

where a and b are loc and scale parameter respectively.

Parameters

  • loc (int, float, list, numpy.ndarray, Tensor) – The location of Gumbel distribution. Default: None.

  • scale (int, float, list, numpy.ndarray, Tensor) – The scale of Gumbel distribution. Default: None.

  • seed (int) – the seed used in sampling. The global seed is used if it is None. Default: 0.

  • dtype (mindspore.dtype) – type of the distribution. Default: mstype.float32.

  • name (str) – the name of the distribution. Default: ‘Gumbel’.

Inputs and Outputs of APIs:

The accessible APIs of the Gumbel distribution are defined in the base class, including:

  • prob, log_prob, cdf, log_cdf, survival_function, and log_survival

  • mean, sd, mode, var, and entropy

  • kl_loss and cross_entropy

  • sample

For more details of all APIs, including the inputs and outputs of all APIs of the Gumbel distribution, please refer to mindspore.nn.probability.distribution.Distribution, and examples below.

Supported Platforms:

Ascend GPU

Note

scale must be greater than zero. dist_spec_args are loc and scale. dtype must be a float type because Gumbel distributions are continuous.

Raises

  • ValueError – When scale <= 0.

  • TypeError – When the input dtype is not a subclass of float.

Examples

>>> import numpy as np
>>> import mindspore
>>> import mindspore.nn as nn
>>> import mindspore.nn.probability.distribution as msd
>>> from mindspore import Tensor
>>> class Prob(nn.Cell):
...     def __init__(self):
...         super(Prob, self).__init__()
...         self.gum = msd.Gumbel(np.array([0.0]), np.array([[1.0], [2.0]]), dtype=mindspore.float32)
...
...     def construct(self, x_):
...         return self.gum.prob(x_)
>>> value = np.array([1.0, 2.0]).astype(np.float32)
>>> pdf = Prob()
>>> output = pdf(Tensor(value, dtype=mindspore.float32))
extend_repr()[source]

Display instance object as string.

property loc

Return the location of the distribution after casting to dtype.

Output:

Tensor, the loc parameter of the distribution.

property scale

Return the scale of the distribution after casting to dtype.

Output:

Tensor, the scale parameter of the distribution.