forked from mindspore-Ecosystem/mindspore
doc update for probability module
fix typo fix pylint problem fix warning fix typo fix typo update formula in log-normal
This commit is contained in:
Zichun Ye
parent
bda20724a9
commit
1f4d906f17
|
|
@ -3,7 +3,7 @@ mindspore.nn.probability.bijector.Bijector
|
|||
|
||||
.. py:class:: mindspore.nn.probability.bijector.Bijector(is_constant_jacobian=False, is_injective=True, name=None, dtype=None, param=None)
|
||||
|
||||
Bijector类。
|
||||
Bijector类。Bijector描述了一种随机变量的映射方法，可以通过一个已有的随机变量 :math:`X` 和一个映射函数 :math:`g` 生成一个新的随机变量 :math:`Y = g(X)` 。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -19,23 +19,22 @@ mindspore.nn.probability.bijector.Bijector
|
|||
|
||||
.. note::
|
||||
Bijector的 `dtype` 为None时，输入值必须是float类型，除此之外没有其他强制要求。在初始化过程中，当 `dtype` 为None时，对参数的数据类型没有强制要求。但所有参数都应具有相同的float类型，否则将引发TypeError。具体来说，参数类型跟随输入值的数据类型，即当 `dtype` 为None时，Bijector的参数将被强制转换为与输入值相同的类型。当指定了 `dtype` 时，参数和输入值的 `dtype` 必须相同。当参数类型或输入值类型与 `dtype` 不相同时，将引发TypeError。只能使用mindspore的float数据类型来指定Bijector的 `dtype` 。
|
||||
|
||||
|
||||
.. py:method:: cast_param_by_value(value, para)
|
||||
|
||||
将Bijector的参数para的数据类型转换为与value相同的类型。
|
||||
|
||||
将输入中的 `para` 的数据类型转换为与 `value` 相同的类型，一般由Bijector的子类用于基于输入对自身参数进行数据类型变化。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入数据。
|
||||
- **para** (Tensor) - Bijector参数。
|
||||
- **kwargs** (dict) - 函数需要的关键字参数字典。
|
||||
|
||||
.. py:method:: construct(name, *args, **kwargs)
|
||||
|
||||
重写Cell中的 `construct` 。
|
||||
|
||||
.. note::
|
||||
支持的函数包括：'forward'、'inverse'、'forward_log_jacobian'、'inverse_log_jacobian'。
|
||||
支持的函数名称包括：'forward'、'inverse'、'forward_log_jacobian'、'inverse_log_jacobian'。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -45,41 +44,56 @@ mindspore.nn.probability.bijector.Bijector
|
|||
|
||||
.. py:method:: forward(value, *args, **kwargs)
|
||||
|
||||
正变换：将输入值转换为另一个分布。
|
||||
正映射，计算输入随机变量 :math:`X = value` 经过映射后的值 :math:`Y = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入。
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
- **args** (list) - 函数所需的位置参数列表。
|
||||
- **kwargs** (dict) - 函数所需的关键字参数字典。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(value, *args, **kwargs)
|
||||
|
||||
对正变换导数取对数。
|
||||
计算正映射导数的对数值，即 :math:`\log(dg(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入。
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
- **args** (list) - 函数所需的位置参数列表。
|
||||
- **kwargs** (dict) - 函数所需的关键字参数字典。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 正映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(value, *args, **kwargs)
|
||||
|
||||
逆变换：将输入值转换回原始分布。
|
||||
逆映射，计算输出随机变量 :math:`Y = value` 时对应的输入随机变量的值 :math:`X = g^{-1}(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入。
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
- **args** (list) - 函数所需的位置参数列表。
|
||||
- **kwargs** (dict) - 函数所需的关键字参数字典。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输入随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(value, *args, **kwargs)
|
||||
|
||||
对逆变换的导数取对数。
|
||||
|
||||
计算逆映射导数的对数值，即 :math:`\log(dg^{-1}(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入。
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
- **args** (list) - 函数所需的位置参数列表。
|
||||
- **kwargs** (dict) - 函数所需的关键字参数字典。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 逆映射导数的对数值。
|
||||
|
|
|
|||
|
|
@ -4,7 +4,7 @@ mindspore.nn.probability.bijector.Exp
|
|||
.. py:class:: mindspore.nn.probability.bijector.Exp(name='Exp')
|
||||
|
||||
指数Bijector（Exponential Bijector）。
|
||||
此Bijector执行如下操作：
|
||||
此Bijector对应的映射函数为：
|
||||
|
||||
.. math::
|
||||
Y = \exp(x).
|
||||
|
|
@ -39,3 +39,51 @@ mindspore.nn.probability.bijector.Exp
|
|||
>>> print(ans4.shape)
|
||||
(3,)
|
||||
|
||||
|
||||
.. py:method:: forward(value)
|
||||
|
||||
正映射，计算输入随机变量 :math:`X = value` 经过映射后的值 :math:`Y = \exp(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(value)
|
||||
|
||||
计算正映射导数的对数值，即 :math:`\log(d\exp(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 正映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(value)
|
||||
|
||||
正映射，计算输出随机变量 :math:`Y = value` 时对应的输入随机变量的值 :math:`X = \log(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输入随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(value)
|
||||
|
||||
计算逆映射导数的对数值，即 :math:`\log(d\log(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 逆映射导数的对数值。
|
||||
|
|
@ -4,10 +4,10 @@ mindspore.nn.probability.bijector.GumbelCDF
|
|||
.. py:class:: mindspore.nn.probability.bijector.GumbelCDF(loc=0.0, scale=1.0, name='GumbelCDF')
|
||||
|
||||
GumbelCDF Bijector。
|
||||
此Bijector执行如下操作：
|
||||
此Bijector对应的映射函数为：
|
||||
|
||||
.. math::
|
||||
Y = \exp(-\exp(\frac{-(X - loc)}{scale}))
|
||||
Y = g(x) = \exp(-\exp(\frac{-(X - loc)}{scale}))
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -20,11 +20,15 @@ mindspore.nn.probability.bijector.GumbelCDF
|
|||
``Ascend`` ``GPU``
|
||||
|
||||
.. note::
|
||||
`scale` 必须大于零。对于 `inverse` 和 `inverse_log_jacobian` ，输入应在(0, 1)范围内。`loc` 和 `scale` 的数据类型必须为float。如果 `loc` 、 `scale` 作为numpy.ndarray或Tensor传入，则它们必须具有相同的数据类型，否则将引发错误。
|
||||
`scale` 中元素必须大于零。对于 `inverse` 和 `inverse_log_jacobian` ，输入应在(0, 1)范围内。`loc` 和 `scale` 中元素的数据类型必须为float。如果 `loc` 、 `scale` 作为numpy.ndarray或Tensor传入，则它们必须具有相同的数据类型，否则将引发错误。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **TypeError** - `loc` 或 `scale` 的数据类型不为float，或 `loc` 和 `scale` 的数据类型不相同。
|
||||
- **TypeError** - `loc` 或 `scale` 中元素的数据类型不为float，或 `loc` 和 `scale` 中元素的数据类型不相同。
|
||||
|
||||
**支持平台：**
|
||||
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -51,3 +55,50 @@ mindspore.nn.probability.bijector.GumbelCDF
|
|||
>>> print(ans4.shape)
|
||||
(3,)
|
||||
|
||||
.. py:method:: forward(value)
|
||||
|
||||
正映射，计算输入随机变量 :math:`X = value` 经过映射后的值 :math:`Y = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输入随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(value)
|
||||
|
||||
计算正映射导数的对数值，即 :math:`\log(dg(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 正映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(value)
|
||||
|
||||
正映射，计算输出随机变量 :math:`Y = value` 时对应的输入随机变量的值 :math:`X = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(value)
|
||||
|
||||
计算逆映射导数的对数值，即 :math:`\log(dg^{-1}(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 逆映射导数的对数值。
|
||||
|
|
|
|||
|
|
@ -3,7 +3,8 @@ mindspore.nn.probability.bijector.Invert
|
|||
|
||||
.. py:class:: mindspore.nn.probability.bijector.Invert(bijector, name='')
|
||||
|
||||
反转Bijector（Invert Bijector），计算输入Bijector的反函数。
|
||||
逆映射Bijector（Invert Bijector）。
|
||||
计算输入Bijector的逆映射。如果正向映射（下面的`bijector`输入)对应的映射函数为 :math:`Y = g(X)`，那么对应的逆映射Bijector的映射函数为 :math:`Y = h(X) = g^{-1}(X)` 。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -38,37 +39,57 @@ mindspore.nn.probability.bijector.Invert
|
|||
.. py:method:: bijector
|
||||
:property:
|
||||
|
||||
返回基础Bijector。
|
||||
Bijector类，返回基础Bijector。
|
||||
|
||||
.. py:method:: forward(x)
|
||||
|
||||
逆变换：将输入值转换回原始分布。
|
||||
计算基础Bijector的逆映射，即 :math:`Y = h(X) = g^{-1}(X)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **x** (Tensor) - 输入。
|
||||
- **x** (Tensor) - 基础Bijector的输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 基础Bijector的输入随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(x)
|
||||
|
||||
逆变换导数的对数。
|
||||
计算基础Bijector的逆映射导数的对数值，即 :math:`\log dg^{-1}(x) / dx`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **x** (Tensor) - 输入。
|
||||
- **x** (Tensor) - 基础Bijector的输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 基类逆映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(y)
|
||||
|
||||
正变换：将输入值转换为另一个分布。
|
||||
计算基础Bijector的正映射，即 :math:`Y = g(X)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **y** (Tensor) - 输入。
|
||||
- **y** (Tensor) - 基础Bijector的输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 基础Bijector的输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(y)
|
||||
|
||||
正变换导数的对数。
|
||||
计算基础Bijector的正映射导数的对数，即 :math:`Y = \log dg(x) / dx`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **y** (Tensor) - 输入。
|
||||
- **y** (Tensor) - 基础Bijector的输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 基类正映射导数的对数值。
|
||||
|
||||
|
|
|
|||
|
|
@ -4,7 +4,7 @@ mindspore.nn.probability.bijector.PowerTransform
|
|||
.. py:class:: mindspore.nn.probability.bijector.PowerTransform(power=0., name='PowerTransform')
|
||||
|
||||
乘方Bijector（Power Bijector）。
|
||||
此Bijector执行如下操作：
|
||||
此Bijector对应的映射函数为：
|
||||
|
||||
.. math::
|
||||
Y = g(X) = (1 + X * c)^{1 / c}, X >= -1 / c
|
||||
|
|
@ -13,7 +13,7 @@ mindspore.nn.probability.bijector.PowerTransform
|
|||
|
||||
Power Bijector将输入从 `[-1/c, inf]` 映射到 `[0, inf]` 。
|
||||
|
||||
当 `c=0` 时，此Bijector等于Exp Bijector。
|
||||
当 `c=0` 时，此Bijector等于 :class:`mindspore.nn.probability.bijector.Exp` Bijector。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -25,12 +25,12 @@ mindspore.nn.probability.bijector.PowerTransform
|
|||
``Ascend`` ``GPU``
|
||||
|
||||
.. note::
|
||||
`power` 的数据类型必须为float。
|
||||
`power` 中元素的数据类型必须为float。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `power` 小于0或静态未知。
|
||||
- **TypeError** - `power` 的数据类型不是float。
|
||||
- **ValueError** - `power` 中元素小于0或静态未知。
|
||||
- **TypeError** - `power` 中元素的数据类型不是float。
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -54,3 +54,50 @@ mindspore.nn.probability.bijector.PowerTransform
|
|||
>>> print(ans4.shape)
|
||||
(3,)
|
||||
|
||||
.. py:method:: forward(value)
|
||||
|
||||
正映射，计算输入随机变量 :math:`X = value` 经过映射后的值 :math:`Y = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输入随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(value)
|
||||
|
||||
计算正映射导数的对数值，即 :math:`\log(dg(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 正映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(value)
|
||||
|
||||
正映射，计算输出随机变量 :math:`Y = value` 时对应的输入随机变量的值 :math:`X = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(value)
|
||||
|
||||
计算逆映射导数的对数值，即 :math:`\log(dg^{-1}(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 逆映射导数的对数值。
|
||||
|
|
|
|||
|
|
@ -4,7 +4,7 @@ mindspore.nn.probability.bijector.ScalarAffine
|
|||
.. py:class:: mindspore.nn.probability.bijector.ScalarAffine(scale=1.0, shift=0.0, name='ScalarAffine')
|
||||
|
||||
标量仿射Bijector（Scalar Affine Bijector）。
|
||||
此Bijector执行如下操作：
|
||||
此Bijector对应的映射函数为：
|
||||
|
||||
.. math::
|
||||
Y = a * X + b
|
||||
|
|
@ -22,11 +22,11 @@ mindspore.nn.probability.bijector.ScalarAffine
|
|||
``Ascend`` ``GPU``
|
||||
|
||||
.. note::
|
||||
`shift` 和 `scale` 的数据类型必须为float。如果 `shift` 、 `scale` 作为numpy.ndarray或Tensor传入，则它们必须具有相同的数据类型，否则将引发错误。
|
||||
`shift` 和 `scale` 中元素的数据类型必须为float。如果 `shift` 、 `scale` 作为numpy.ndarray或Tensor传入，则它们必须具有相同的数据类型，否则将引发错误。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **TypeError** - `shift` 或 `scale` 的数据类型不为float，或 `shift` 和 `scale` 的数据类型不相同。
|
||||
- **TypeError** - `shift` 或 `scale` 中元素的数据类型不为float，或 `shift` 和 `scale` 的数据类型不相同。
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -50,3 +50,50 @@ mindspore.nn.probability.bijector.ScalarAffine
|
|||
>>> print(ans4.shape)
|
||||
()
|
||||
|
||||
.. py:method:: forward(value)
|
||||
|
||||
正映射，计算输入随机变量 :math:`X = value` 经过映射后的值 :math:`Y = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输入随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(value)
|
||||
|
||||
计算正映射导数的对数值，即 :math:`\log(dg(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 正映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(value)
|
||||
|
||||
正映射，计算输出随机变量 :math:`Y = value` 时对应的输入随机变量的值 :math:`X = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(value)
|
||||
|
||||
计算逆映射导数的对数值，即 :math:`\log(dg^{-1}(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 逆映射导数的对数值。
|
||||
|
|
|
|||
|
|
@ -4,10 +4,10 @@ mindspore.nn.probability.bijector.Softplus
|
|||
.. py:class:: mindspore.nn.probability.bijector.Softplus(sharpness=1.0, name='Softplus')
|
||||
|
||||
Softplus Bijector。
|
||||
此Bijector执行如下操作：
|
||||
此Bijector对应的映射函数为：
|
||||
|
||||
.. math::
|
||||
Y = \frac{\log(1 + e ^ {kX})}{k}
|
||||
Y = g(x) = \frac{\log(1 + e ^ {kX})}{k}
|
||||
|
||||
其中k是锐度因子。
|
||||
|
||||
|
|
@ -21,11 +21,11 @@ mindspore.nn.probability.bijector.Softplus
|
|||
``Ascend`` ``GPU``
|
||||
|
||||
.. note::
|
||||
`sharpness` 的数据类型必须为float。
|
||||
`sharpness` 中元素的数据类型必须为float。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **TypeError** - sharpness的数据类型不为float。
|
||||
- **TypeError** - sharpness中元素的数据类型不为float。
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -51,3 +51,50 @@ mindspore.nn.probability.bijector.Softplus
|
|||
>>> print(ans4.shape)
|
||||
(3,)
|
||||
|
||||
.. py:method:: forward(value)
|
||||
|
||||
正映射，计算输入随机变量 :math:`X = value` 经过映射后的值 :math:`Y = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输入随机变量的值。
|
||||
|
||||
.. py:method:: forward_log_jacobian(value)
|
||||
|
||||
计算正映射导数的对数值，即 :math:`\log(dg(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输入随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 正映射导数的对数值。
|
||||
|
||||
.. py:method:: inverse(value)
|
||||
|
||||
正映射，计算输出随机变量 :math:`Y = value` 时对应的输入随机变量的值 :math:`X = g(value)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 输出随机变量的值。
|
||||
|
||||
.. py:method:: inverse_log_jacobian(value)
|
||||
|
||||
计算逆映射导数的对数值，即 :math:`\log(dg^{-1}(x) / dx)`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 输出随机变量的值。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 逆映射导数的对数值。
|
||||
|
|
|
|||
|
|
@ -4,6 +4,7 @@ mindspore.nn.probability.distribution.Bernoulli
|
|||
.. py:class:: mindspore.nn.probability.distribution.Bernoulli(probs=None, seed=None, dtype=mstype.int32, name='Bernoulli')
|
||||
|
||||
伯努利分布（Bernoulli Distribution）。
|
||||
离散随机分布，取值范围为 :math:`\{0, 1\}` ，概率质量函数为 :math:`P(X = 0) = p, P(X = 1) = 1-p`。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -17,7 +18,12 @@ mindspore.nn.probability.distribution.Bernoulli
|
|||
``Ascend`` ``GPU``
|
||||
|
||||
.. note::
|
||||
`probs` 必须是合适的概率（0<p<1）。
|
||||
`probs` 中元素必须是合适的概率（0<p<1）。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `probs` 中元素小于0或大于1。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -104,4 +110,8 @@ mindspore.nn.probability.distribution.Bernoulli
|
|||
|
||||
.. py:method:: probs
|
||||
|
||||
返回结果为1的概率。
|
||||
返回结果为1的概率。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 结果为1的概率。
|
||||
|
|
|
|||
|
|
@ -3,12 +3,18 @@ mindspore.nn.probability.distribution.Beta
|
|||
|
||||
.. py:class:: mindspore.nn.probability.distribution.Beta(concentration1=None, concentration0=None, seed=None, dtype=mstype.float32, name='Beta')
|
||||
|
||||
贝塔分布（Beta Distribution）。
|
||||
Beta 分布（Beta Distribution）。
|
||||
连续随机分布，取值范围为 :math:`[0, 1]` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, \alpha, \beta) = x^\alpha (1-x)^{\beta - 1} / B(\alpha, \beta).
|
||||
|
||||
其中 :math:`B` 为 Beta 函数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **concentration1** (list, numpy.ndarray, Tensor) - 贝塔分布的alpha。
|
||||
- **concentration0** (list, numpy.ndarray, Tensor) - 贝塔分布的beta。
|
||||
- **concentration1** (int, float, list, numpy.ndarray, Tensor) - Beta 分布的alpha。
|
||||
- **concentration0** (int, float, list, numpy.ndarray, Tensor) - Beta 分布的beta。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 采样结果的数据类型。默认值：mindspore.float32。
|
||||
- **name** (str) - 分布的名称。默认值：'Beta'。
|
||||
|
|
@ -18,8 +24,13 @@ mindspore.nn.probability.distribution.Beta
|
|||
``Ascend``
|
||||
|
||||
.. note::
|
||||
- `concentration1` 和 `concentration0` 必须大于零。
|
||||
- `dtype` 必须是float，因为贝塔分布是连续的。
|
||||
- `concentration1` 和 `concentration0` 中元素必须大于零。
|
||||
- `dtype` 必须是float，因为 Beta 分布是连续的。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `concentration1` 或者 `concentration0` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -28,7 +39,7 @@ mindspore.nn.probability.distribution.Beta
|
|||
>>> import mindspore.nn.probability.distribution as msd
|
||||
>>> from mindspore import Tensor
|
||||
>>>
|
||||
>>> # 初始化concentration1为3.0和concentration0为4.0的贝塔分布。
|
||||
>>> # 初始化concentration1为3.0和concentration0为4.0的 Beta 分布。
|
||||
>>> b1 = msd.Beta([3.0], [4.0], dtype=mindspore.float32)
|
||||
>>>
|
||||
>>> # Beta分布可以在没有参数的情况下初始化。
|
||||
|
|
@ -113,10 +124,18 @@ mindspore.nn.probability.distribution.Beta
|
|||
.. py:method:: concentration0
|
||||
:property:
|
||||
|
||||
返回concentration0（也称为贝塔分布的beta）。
|
||||
返回concentration0（也称为 Beta 分布的beta）。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, concentration0 的值。
|
||||
|
||||
.. py:method:: concentration1
|
||||
:property:
|
||||
|
||||
返回concentration1（也称为贝塔分布的alpha）。
|
||||
返回concentration1（也称为 Beta 分布的alpha）。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, concentration1 的值。
|
||||
|
||||
|
|
|
|||
|
|
@ -4,6 +4,7 @@ mindspore.nn.probability.distribution.Categorical
|
|||
.. py:class:: mindspore.nn.probability.distribution.Categorical(probs=None, seed=None, dtype=mstype.float32, name='Categorical')
|
||||
|
||||
分类分布。
|
||||
离散随机分布，取值范围为 :math:`\{1, 2, ..., k\}` ，概率质量函数为 :math:`P(X = i) = p_i, i = 1, ..., k`。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -19,6 +20,10 @@ mindspore.nn.probability.distribution.Categorical
|
|||
.. note::
|
||||
`probs` 的秩必须至少为1，值是合适的概率，并且总和为1。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `probs` 的秩为0或者其中所有元素的和不等于1。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -81,8 +86,12 @@ mindspore.nn.probability.distribution.Categorical
|
|||
>>> ans = ca2.kl_loss('Categorical', probs_b, probs_a)
|
||||
>>> print(ans.shape)
|
||||
()
|
||||
|
||||
|
||||
.. py:method:: probs
|
||||
|
||||
返回事件概率。
|
||||
|
||||
返回事件发生的概率。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 事件发生的概率。
|
||||
|
||||
|
|
|
|||
|
|
@ -4,6 +4,12 @@ mindspore.nn.probability.distribution.Cauchy
|
|||
.. py:class:: mindspore.nn.probability.distribution.Cauchy(loc=None, scale=None, seed=None, dtype=mstype.float32, name='Cauchy')
|
||||
|
||||
柯西分布（Cauchy distribution）。
|
||||
连续随机分布，取值范围为所有实数，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / \pi b(1 - ((x - a)/b)^2).
|
||||
|
||||
其中 :math:`a, b` 为分别为柯西分布的位置参数和比例参数。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -18,10 +24,16 @@ mindspore.nn.probability.distribution.Cauchy
|
|||
``Ascend``
|
||||
|
||||
.. note::
|
||||
- `scale` 必须大于零。
|
||||
- `scale` 中的元素必须大于零。
|
||||
- `dtype` 必须是float，因为柯西分布是连续的。
|
||||
- GPU后端不支持柯西分布。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `scale` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -107,14 +119,21 @@ mindspore.nn.probability.distribution.Cauchy
|
|||
>>> ans = cauchy2.sample((2,3), loc_a, scale_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 3)
|
||||
|
||||
|
||||
.. py:method:: loc
|
||||
:property:
|
||||
|
||||
返回分布位置。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的位置值。
|
||||
|
||||
.. py:method:: scale
|
||||
:property:
|
||||
|
||||
返回分布比例。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的比例值。
|
||||
|
|
|
|||
|
|
@ -18,20 +18,24 @@ mindspore.nn.probability.distribution.Distribution
|
|||
|
||||
.. note::
|
||||
派生类必须重写 `_mean` 、 `_prob` 和 `_log_prob` 等操作。必填参数必须通过 `args` 或 `kwargs` 传入，如 `_prob` 的 `value` 。
|
||||
|
||||
|
||||
.. py:method:: cdf(value, *args, **kwargs)
|
||||
|
||||
在给定值下评估累积分布函数（Cumulatuve Distribution Function, CDF）。
|
||||
在给定值下计算累积分布函数（Cumulatuve Distribution Function, CDF）。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 要评估的值。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
|
||||
- **value** (Tensor) - 要计算的值。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 累积分布函数的值。
|
||||
|
||||
.. py:method:: construct(name, *args, **kwargs)
|
||||
|
||||
重写Cell中的 `construct` 。
|
||||
|
|
@ -45,137 +49,185 @@ mindspore.nn.probability.distribution.Distribution
|
|||
- **name** (str) - 函数名称。
|
||||
- **args** (list) - 函数所需的位置参数列表。
|
||||
- **kwargs** (dict) - 函数所需的关键字参数字典。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, name对应函数的值。
|
||||
|
||||
.. py:method:: cross_entropy(dist, *args, **kwargs)
|
||||
|
||||
评估分布a和b之间的交叉熵。
|
||||
计算分布a和b之间的交叉熵。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **dist** (str) - 分布的类型。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
Distribution b的 `dist_spec_args` 必须通过 `args` 或 `kwargs` 传递给函数。 传入Distribution a的 `dist_spec_args` 是可选的。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 交叉熵的值。
|
||||
|
||||
.. py:method:: entropy(*args, **kwargs)
|
||||
|
||||
计算熵。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 熵的值。
|
||||
|
||||
.. py:method:: get_dist_args(*args, **kwargs)
|
||||
|
||||
检查默认参数的可用性和有效性。
|
||||
返回分布的参数列表。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
传递给字类的参数的顺序应该与通过 `_add_parameter` 初始化默认参数的顺序相同。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
list[Tensor], 参数列表。
|
||||
|
||||
.. py:method:: get_dist_type()
|
||||
|
||||
返回分布类型。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
string, 分布类型名字。
|
||||
|
||||
.. py:method:: kl_loss(dist, *args, **kwargs)
|
||||
|
||||
评估KL散度，即KL(a||b)。
|
||||
计算KL散度，即KL(a||b)。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **dist** (str) - 分布的类型。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
Distribution b的 `dist_spec_args` 必须通过 `args` 或 `kwargs` 传递给函数。 传入Distribution a的 `dist_spec_args` 是可选的。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, KL散度。
|
||||
|
||||
.. py:method:: log_cdf(value, *args, **kwargs)
|
||||
|
||||
计算给定值对于的cdf的对数。
|
||||
计算给定值对于的累积分布函数的对数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 要评估的值。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **value** (Tensor) - 要计算的值。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 累积分布函数的对数。
|
||||
|
||||
.. py:method:: log_prob(value, *args, **kwargs)
|
||||
|
||||
计算给定值对应的概率的对数（pdf或pmf）。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 要评估的值。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **value** (Tensor) - 要计算的值。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 累积分布函数的对数。
|
||||
|
||||
.. py:method:: log_survival(value, *args, **kwargs)
|
||||
|
||||
计算给定值对应的剩余函数的对数。
|
||||
计算给定值对应的生存函数的对数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 要评估的值。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **value** (Tensor) - 要计算的值。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 生存函数的对数。
|
||||
|
||||
.. py:method:: mean(*args, **kwargs)
|
||||
|
||||
评估平均值。
|
||||
计算期望。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 概率分布的期望。
|
||||
|
||||
.. py:method:: mode(*args, **kwargs)
|
||||
|
||||
评估模式。
|
||||
计算众数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 概率分布的众数。
|
||||
|
||||
.. py:method:: prob(value, *args, **kwargs)
|
||||
|
||||
评估给定值下的概率（Probability Density Function或Probability Mass Function）。
|
||||
计算给定值下的概率。对于离散分布是计算概率质量函数（Probability Mass Function），而对于连续分布是计算概率密度函数（Probability Density Function）。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 要评估的值。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
|
||||
- **value** (Tensor) - 要计算的值。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 概率值。
|
||||
|
||||
.. py:method:: sample(*args, **kwargs)
|
||||
|
||||
采样函数。
|
||||
|
|
@ -183,46 +235,62 @@ mindspore.nn.probability.distribution.Distribution
|
|||
**参数：**
|
||||
|
||||
- **shape** (tuple) - 样本的shape。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 根据概率分布采样的样本。
|
||||
|
||||
.. py:method:: sd(*args, **kwargs)
|
||||
|
||||
标准差评估。
|
||||
计算标准差。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 概率分布的标准差。
|
||||
|
||||
.. py:method:: survival_function(value, *args, **kwargs)
|
||||
|
||||
计算给定值对应的剩余函数。
|
||||
计算给定值对应的生存函数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **value** (Tensor) - 要评估的值。
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
|
||||
- **value** (Tensor) - 要计算的值。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 生存函数的值。
|
||||
|
||||
.. py:method:: var(*args, **kwargs)
|
||||
|
||||
评估方差。
|
||||
计算方差。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **args** (list) - 传递给子类的位置参数列表。
|
||||
- **kwargs** (dict) - 传递给子类的关键字参数字典。
|
||||
- **args** (list) - 位置参数列表，具体需要的参数根据子类的实现确定。。
|
||||
- **kwargs** (dict) - 关键字参数字典，具体需要的参数根据子类的实现确定。
|
||||
|
||||
.. note::
|
||||
可以通过 `args` 或 `kwargs` 传递其 `dist_spec_args` 来选择性地将Distribution传递给函数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 概率分布的方差。
|
||||
|
||||
|
|
|
|||
|
|
@ -3,11 +3,17 @@ mindspore.nn.probability.distribution.Exponential
|
|||
|
||||
.. py:class:: mindspore.nn.probability.distribution.Exponential(rate=None, seed=None, dtype=mstype.float32, name='Exponential')
|
||||
|
||||
示例类：指数分布（Exponential Distribution）。
|
||||
指数分布（Exponential Distribution）。
|
||||
连续随机分布，取值范围为所有实数，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, \lambda) = \lambda \exp(-\lambda x).
|
||||
|
||||
其中 :math:`\lambda` 为分别为指数分布的率参数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **rate** (float, list, numpy.ndarray, Tensor) - 逆指数。默认值：None。
|
||||
- **rate** (int, float, list, numpy.ndarray, Tensor) - 率参数。默认值：None。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 事件样例的类型。默认值：mindspore.float32。
|
||||
- **name** (str) - 分布的名称。默认值：'Exponential'。
|
||||
|
|
@ -17,9 +23,14 @@ mindspore.nn.probability.distribution.Exponential
|
|||
``Ascend`` ``GPU``
|
||||
|
||||
.. note::
|
||||
- `rate` 必须大于0。
|
||||
- `rate` 中的元素必须大于0。
|
||||
- `dtype` 必须是float，因为指数分布是连续的。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `rate` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -98,9 +109,13 @@ mindspore.nn.probability.distribution.Exponential
|
|||
>>> ans = e2.sample((2,3), rate_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 1)
|
||||
|
||||
|
||||
.. py:method:: rate
|
||||
:property:
|
||||
|
||||
返回 `rate` 。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, rate 的值。
|
||||
|
||||
|
|
|
|||
|
|
@ -4,11 +4,17 @@ mindspore.nn.probability.distribution.Gamma
|
|||
.. py:class:: mindspore.nn.probability.distribution.Gamma(concentration=None, rate=None, seed=None, dtype=mstype.float3, name='Gamma')
|
||||
|
||||
伽马分布（Gamma distribution）。
|
||||
连续随机分布，取值范围为 :math:`(0, \inf)` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, \alpha, \beta) = \beta^\alpha / \Gamma(\alpha) x^{\alpha - 1} \exp(-\beta x).
|
||||
|
||||
其中 :math:`G` 为 Gamma 函数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **concentration** (list, numpy.ndarray, Tensor) - 浓度，也被称为伽马分布的alpha。默认值：None。
|
||||
- **rate** (list, numpy.ndarray, Tensor) - 逆尺度参数，也被称为伽马分布的beta。默认值：None。
|
||||
- **concentration** (int, float, list, numpy.ndarray, Tensor) - 浓度，也被称为伽马分布的alpha。默认值：None。
|
||||
- **rate** (int, float, list, numpy.ndarray, Tensor) - 逆尺度参数，也被称为伽马分布的beta。默认值：None。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 事件样例的类型。默认值：mindspore.float32。
|
||||
- **name** (str) - 分布的名称。默认值：'Gamma'。
|
||||
|
|
@ -18,9 +24,14 @@ mindspore.nn.probability.distribution.Gamma
|
|||
``Ascend``
|
||||
|
||||
.. note::
|
||||
- `concentration` 和 `rate` 必须大于零。
|
||||
- `concentration` 和 `rate` 中的元素必须大于零。
|
||||
- `dtype` 必须是float，因为伽马分布是连续的。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `concentration` 或者 `rate` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -107,14 +118,22 @@ mindspore.nn.probability.distribution.Gamma
|
|||
>>> ans = g2.sample((2,3), concentration_a, rate_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 3)
|
||||
|
||||
|
||||
.. py:method:: concentration
|
||||
:property:
|
||||
|
||||
返回分布的浓度（也称为伽马分布的alpha）。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, concentration 的值。
|
||||
|
||||
.. py:method:: rate
|
||||
:property:
|
||||
|
||||
返回分布的逆尺度（也称为伽马分布的beta）。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, rate 的值。
|
||||
|
||||
|
|
|
|||
|
|
@ -6,10 +6,11 @@ mindspore.nn.probability.distribution.Geometric
|
|||
几何分布（Geometric Distribution）。
|
||||
|
||||
它代表在第一次成功之前有k次失败，即在第一次成功实现时，总共有k+1个伯努利试验。
|
||||
离散随机分布，取值范围为正自然数集，概率质量函数为 :math:`P(X = i) = p(1-p)^{i-1}, i = 1, 2, ...`。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **probs** (float, list, numpy.ndarray, Tensor) - 成功的概率。默认值：None。
|
||||
- **probs** (int, float, list, numpy.ndarray, Tensor) - 成功的概率。默认值：None。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 事件样例的类型。默认值：mindspore.int32.
|
||||
- **name** (str) - 分布的名称。默认值：'Geometric'。
|
||||
|
|
@ -21,6 +22,11 @@ mindspore.nn.probability.distribution.Geometric
|
|||
.. note::
|
||||
`probs` 必须是合适的概率（0<p<1）。
|
||||
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `probs` 中元素小于0或者大于1。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -101,9 +107,13 @@ mindspore.nn.probability.distribution.Geometric
|
|||
>>> ans = g2.sample((2,3), probs_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 1)
|
||||
|
||||
|
||||
.. py:method:: probs
|
||||
:property:
|
||||
|
||||
返回伯努利试验成功的概率。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 伯努利试验成功的概率值。
|
||||
|
||||
|
|
|
|||
|
|
@ -3,12 +3,18 @@ mindspore.nn.probability.distribution.Gumbel
|
|||
|
||||
.. py:class:: mindspore.nn.probability.distribution.Gumbel(loc, scale, seed=0, dtype=mstype.float32, name='Gumbel')
|
||||
|
||||
耿贝尔分布（Gumbel distribution）。
|
||||
Gumbel分布（Gumbel distribution）。
|
||||
连续随机分布，取值范围为 :math:`(0, \inf)` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / b \exp(\exp(-(x - a) / b) - x).
|
||||
|
||||
其中 :math:`a, b` 为分别为Gumbel分布的位置参数和比例参数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **loc** (float, list, numpy.ndarray, Tensor) - 耿贝尔分布的位置。
|
||||
- **scale** (float, list, numpy.ndarray, Tensor) - 耿贝尔分布的尺度。
|
||||
- **loc** (int, float, list, numpy.ndarray, Tensor) - Gumbel分布的位置。
|
||||
- **scale** (int, float, list, numpy.ndarray, Tensor) - Gumbel分布的尺度。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 分布类型。默认值：mindspore.float32。
|
||||
- **name** (str) - 分布的名称。默认值：'Gumbel'。
|
||||
|
|
@ -19,9 +25,14 @@ mindspore.nn.probability.distribution.Gumbel
|
|||
|
||||
.. note::
|
||||
- `scale` 必须大于零。
|
||||
- `dtype` 必须是浮点类型，因为耿贝尔分布是连续的。
|
||||
- `dtype` 必须是浮点类型，因为Gumbel分布是连续的。
|
||||
- GPU后端不支持 `kl_loss` 和 `cross_entropy` 。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `scale` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -38,14 +49,21 @@ mindspore.nn.probability.distribution.Gumbel
|
|||
>>> value = np.array([1.0, 2.0]).astype(np.float32)
|
||||
>>> pdf = Prob()
|
||||
>>> output = pdf(Tensor(value, dtype=mindspore.float32))
|
||||
|
||||
|
||||
.. py:method:: loc
|
||||
:property:
|
||||
|
||||
返回分布位置。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的位置值。
|
||||
|
||||
.. py:method:: scale
|
||||
:property:
|
||||
|
||||
返回分布尺度。
|
||||
|
||||
返回分布比例。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的比例值。
|
||||
|
|
|
|||
|
|
@ -4,7 +4,13 @@ mindspore.nn.probability.distribution.LogNormal
|
|||
.. py:class:: mindspore.nn.probability.distribution.LogNormal(loc=None, scale=None, seed=0, dtype=mstype.float32, name='LogNormal')
|
||||
|
||||
对数正态分布（LogNormal distribution）。
|
||||
对数正态分布是随机变量的连续概率分布，变量的对数为正态分布。它被构造为正态分布的指数变换。
|
||||
连续随机分布，取值范围为 :math:`(0, \inf)` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / xb\sqrt{2\pi} \exp(-(\ln(x) - a)^2 / 2b^2).
|
||||
|
||||
其中 :math:`a, b` 为分别为基础正态分布的平均值和标准差。
|
||||
服从对数正态分布的随机变量的对数服从正态分布。它被构造为正态分布的指数变换。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -22,6 +28,11 @@ mindspore.nn.probability.distribution.LogNormal
|
|||
- `scale` 必须大于零。
|
||||
- `dtype` 必须是float，因为对数正态分布是连续的。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `scale` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import numpy as np
|
||||
|
|
@ -39,13 +50,21 @@ mindspore.nn.probability.distribution.LogNormal
|
|||
>>> output = pdf(Tensor([1.0, 2.0], dtype=mindspore.float32))
|
||||
>>> print(output.shape)
|
||||
(2, 2)
|
||||
|
||||
|
||||
.. py:method:: loc
|
||||
:property:
|
||||
|
||||
返回分布的均值。
|
||||
|
||||
返回分布位置。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的位置值。
|
||||
|
||||
.. py:method:: scale
|
||||
:property:
|
||||
|
||||
返回分布的标准差。
|
||||
返回分布比例。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的比例值。
|
||||
|
|
|
|||
|
|
@ -3,12 +3,18 @@ mindspore.nn.probability.distribution.Logistic
|
|||
|
||||
.. py:class:: mindspore.nn.probability.distribution.Logistic(loc=None, scale=None, seed=None, dtype=mstype.float32, name='Logistic')
|
||||
|
||||
逻辑斯谛分布（Logistic distribution）。
|
||||
Logistic分布（Logistic distribution）。
|
||||
连续随机分布，取值范围为 :math:`(0, \inf)` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / b \exp(\exp(-(x - a) / b) - x).
|
||||
|
||||
其中 :math:`a, b` 为分别为Logistic分布的位置参数和比例参数。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **loc** (int, float, list, numpy.ndarray, Tensor) - 逻辑斯谛分布的位置。默认值：None。
|
||||
- **scale** (int, float, list, numpy.ndarray, Tensor) - 逻辑斯谛分布的尺度。默认值：None。
|
||||
- **loc** (int, float, list, numpy.ndarray, Tensor) - Logistic分布的位置。默认值：None。
|
||||
- **scale** (int, float, list, numpy.ndarray, Tensor) - Logistic分布的尺度。默认值：None。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 事件样例的类型。默认值：mindspore.float32。
|
||||
- **name** (str) - 分布的名称。默认值：'Logistic'。
|
||||
|
|
@ -19,7 +25,12 @@ mindspore.nn.probability.distribution.Logistic
|
|||
|
||||
.. note::
|
||||
- `scale` 必须大于零。
|
||||
- `dtype` 必须是float，因为逻辑斯谛分布是连续的。
|
||||
- `dtype` 必须是float，因为Logistic分布是连续的。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `scale` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
|
|
@ -27,9 +38,9 @@ mindspore.nn.probability.distribution.Logistic
|
|||
>>> import mindspore.nn as nn
|
||||
>>> import mindspore.nn.probability.distribution as msd
|
||||
>>> from mindspore import Tensor
|
||||
>>> # 初始化loc为3.0和scale为4.0的逻辑斯谛分布。
|
||||
>>> # 初始化loc为3.0和scale为4.0的Logistic分布。
|
||||
>>> l1 = msd.Logistic(3.0, 4.0, dtype=mindspore.float32)
|
||||
>>> # 可以在没有参数的情况下初始化逻辑斯谛分布。
|
||||
>>> # 可以在没有参数的情况下初始化Logistic分布。
|
||||
>>> # 在这种情况下，`loc`和`scale`必须通过参数传入。
|
||||
>>> l2 = msd.Logistic(dtype=mindspore.float32)
|
||||
>>>
|
||||
|
|
@ -90,14 +101,22 @@ mindspore.nn.probability.distribution.Logistic
|
|||
>>> ans = l1.sample((2,3), loc_a, scale_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 3)
|
||||
|
||||
|
||||
.. py:method:: loc
|
||||
:property:
|
||||
|
||||
返回分布位置。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的位置值。
|
||||
|
||||
.. py:method:: scale
|
||||
:property:
|
||||
|
||||
返回分布尺度。
|
||||
|
||||
返回分布比例。
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的比例值。
|
||||
|
||||
|
|
|
|||
|
|
@ -4,6 +4,12 @@ mindspore.nn.probability.distribution.Normal
|
|||
.. py:class:: mindspore.nn.probability.distribution.Normal(mean=None, sd=None, seed=None, dtype=mstype.float32, name='Normal')
|
||||
|
||||
正态分布（Normal distribution）。
|
||||
连续随机分布，取值范围为 :math:`(-\inf, \inf)` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, \mu, \sigma) = 1 / \sigma\sqrt{2\pi} \exp(-(x - \mu)^2 / 2\sigma^2).
|
||||
|
||||
其中 :math:`\mu, \sigma` 为分别为正态分布的期望与标准差。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -21,6 +27,11 @@ mindspore.nn.probability.distribution.Normal
|
|||
- `sd` 必须大于零。
|
||||
- `dtype` 必须是float，因为正态分布是连续的。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `sd` 中元素小于0。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
|
|||
|
|
@ -4,10 +4,16 @@ mindspore.nn.probability.distribution.Poisson
|
|||
.. py:class:: mindspore.nn.probability.distribution.Poisson(rate=None, seed=None, dtype=mstype.float32, name='Poisson')
|
||||
|
||||
泊松分布（Poisson Distribution）。
|
||||
离散随机分布，取值范围为正自然数集，概率质量函数为
|
||||
|
||||
.. math::
|
||||
P(X = k) = \lambda^k \exp(-\lambda) / k!, k = 1, 2, ...
|
||||
|
||||
其中 :math:`\lambda` 为率参数（rate)。
|
||||
|
||||
**参数：**
|
||||
|
||||
- **rate** (list, numpy.ndarray, Tensor) - 泊松分布的率参数。默认值：None。
|
||||
- **rate** (int, float, list, numpy.ndarray, Tensor) - 泊松分布的率参数。默认值：None。
|
||||
- **seed** (int) - 采样时使用的种子。如果为None，则使用全局种子。默认值：None。
|
||||
- **dtype** (mindspore.dtype) - 事件样例的类型。默认值：mindspore.float32。
|
||||
- **name** (str) - 分布的名称。默认值：'Poisson'。
|
||||
|
|
@ -19,6 +25,11 @@ mindspore.nn.probability.distribution.Poisson
|
|||
.. note::
|
||||
`rate` 必须大于0。
|
||||
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `rate` 中元素小于0。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -83,9 +94,13 @@ mindspore.nn.probability.distribution.Poisson
|
|||
>>> ans = p2.sample((2,3), rate_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 1)
|
||||
|
||||
|
||||
.. py:method:: rate
|
||||
:property:
|
||||
|
||||
返回分布的 `rate` 参数。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, rate 参数的值。
|
||||
|
||||
|
|
|
|||
|
|
@ -4,7 +4,8 @@ mindspore.nn.probability.distribution.TransformedDistribution
|
|||
.. py:class:: mindspore.nn.probability.distribution.TransformedDistribution(bijector, distribution, seed=None, name='transformed_distribution')
|
||||
|
||||
转换分布（Transformed Distribution）。
|
||||
该类包含一个Bijector和一个分布，并通过Bijector定义的操作将原始分布转换为新分布。
|
||||
该类包含一个Bijector和一个分布，并通过Bijector定义的操作将原始分布转换为新分布。可如果原始分布为 :math:`X` ，Bijector的映射函数为 :math:`g`， 那么对应的转换分布为 :math:`Y = g(X)` 。
|
||||
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -20,6 +21,10 @@ mindspore.nn.probability.distribution.TransformedDistribution
|
|||
.. note::
|
||||
用于初始化原始分布的参数不能为None。例如，由于未指定 `mean` 和 `sd` ，因此无法使用mynormal = msd.Normal(dtype=mindspore.float32)初始化TransformedDistribution。
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - bijector不是Bijector类，distribution不是Distribution类。
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import numpy as np
|
||||
|
|
@ -48,4 +53,4 @@ mindspore.nn.probability.distribution.TransformedDistribution
|
|||
>>> cdf, sample = net(tx)
|
||||
>>> print(sample.shape)
|
||||
(2, 3)
|
||||
|
||||
|
||||
|
|
|
|||
|
|
@ -3,7 +3,13 @@ mindspore.nn.probability.distribution.Uniform
|
|||
|
||||
.. py:class:: mindspore.nn.probability.distribution.Uniform(low=None, high=None, seed=None, dtype=mstype.float32, name='Uniform')
|
||||
|
||||
示例类：均匀分布（Uniform Distribution）。
|
||||
均匀分布（Uniform Distribution）。
|
||||
连续随机分布，取值范围为 :math:`[a, b]` ，概率密度函数为
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / (b - a).
|
||||
|
||||
其中 :math:`a, b` 为分别为均匀分布的下界和上界。
|
||||
|
||||
**参数：**
|
||||
|
||||
|
|
@ -21,6 +27,13 @@ mindspore.nn.probability.distribution.Uniform
|
|||
- `low` 必须小于 `high` 。
|
||||
- `dtype` 必须是float类型，因为均匀分布是连续的。
|
||||
|
||||
|
||||
**异常：**
|
||||
|
||||
- **ValueError** - `low` 大于等于 `high` 。
|
||||
- **TypeError** - `dtype` 不是float的子类。
|
||||
|
||||
|
||||
**样例：**
|
||||
|
||||
>>> import mindspore
|
||||
|
|
@ -108,14 +121,21 @@ mindspore.nn.probability.distribution.Uniform
|
|||
>>> ans = u2.sample((2,3), low_a, high_a)
|
||||
>>> print(ans.shape)
|
||||
(2, 3, 2)
|
||||
|
||||
|
||||
.. py:method:: high
|
||||
:property:
|
||||
|
||||
返回分布的上限。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的上限值。
|
||||
|
||||
.. py:method:: low
|
||||
:property:
|
||||
|
||||
返回分布的下限。
|
||||
|
||||
|
||||
**返回：**
|
||||
|
||||
Tensor, 分布的下限值。
|
||||
|
|
@ -26,7 +26,10 @@ from ..distribution import TransformedDistribution
|
|||
|
||||
class Bijector(Cell):
|
||||
"""
|
||||
Bijecotr class.
|
||||
Bijecotr class. A bijector perform a mapping from one distribution to the other via some function.
|
||||
If X is a random variable following the original distribution,
|
||||
and g(x) is the mapping function,
|
||||
then Y = g(X) is the random variable following the transformed distribution.
|
||||
|
||||
Args:
|
||||
is_constant_jacobian (bool): Whether the Bijector has constant derivative. Default: False.
|
||||
|
|
@ -234,24 +237,56 @@ class Bijector(Cell):
|
|||
def forward(self, value, *args, **kwargs):
|
||||
"""
|
||||
Forward transformation: transform the input value to another distribution.
|
||||
|
||||
Args:
|
||||
value (Tensor): the value of the input variables.
|
||||
*args (list): the list of positional arguments forwarded to subclasses.
|
||||
**kwargs (dict): the dictionary of keyword arguments forwarded to subclasses.
|
||||
|
||||
Output:
|
||||
Tensor, the value of the transformed random variable.
|
||||
"""
|
||||
return self._forward(value, *args, **kwargs)
|
||||
|
||||
def inverse(self, value, *args, **kwargs):
|
||||
"""
|
||||
Inverse transformation: transform the input value back to the original distribution.
|
||||
|
||||
Args:
|
||||
value (Tensor): the value of the transformed variables.
|
||||
*args (list): the list of positional arguments forwarded to subclasses.
|
||||
**kwargs (dict): the dictionary of keyword arguments forwarded to subclasses.
|
||||
|
||||
Output:
|
||||
Tensor, the value of the input random variable.
|
||||
"""
|
||||
return self._inverse(value, *args, **kwargs)
|
||||
|
||||
def forward_log_jacobian(self, value, *args, **kwargs):
|
||||
"""
|
||||
Logarithm of the derivative of the forward transformation.
|
||||
|
||||
Args:
|
||||
value (Tensor): the value of the input variables.
|
||||
*args (list): the list of positional arguments forwarded to subclasses.
|
||||
**kwargs (dict): the dictionary of keyword arguments forwarded to subclasses.
|
||||
|
||||
Output:
|
||||
Tensor, the value of logarithm of the derivative of the forward transformation.
|
||||
"""
|
||||
return self._forward_log_jacobian(value, *args, **kwargs)
|
||||
|
||||
def inverse_log_jacobian(self, value, *args, **kwargs):
|
||||
"""
|
||||
Logarithm of the derivative of the inverse transformation.
|
||||
|
||||
Args:
|
||||
value (Tensor): the value of the transformed variables.
|
||||
*args (list): the list of positional arguments forwarded to subclasses.
|
||||
**kwargs (dict): the dictionary of keyword arguments forwarded to subclasses.
|
||||
|
||||
Output:
|
||||
Tensor, the value of logarithm of the derivative of the inverse transformation.
|
||||
"""
|
||||
return self._inverse_log_jacobian(value, *args, **kwargs)
|
||||
|
||||
|
|
|
|||
|
|
@ -27,6 +27,18 @@ class Exp(PowerTransform):
|
|||
Args:
|
||||
name (str): The name of the Bijector. Default: 'Exp'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- **forward**
|
||||
- **inverse**
|
||||
- **forward_log_jacobian**
|
||||
- **backward_log_jacobian**
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Bijector`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
|
|||
|
|
@ -32,6 +32,19 @@ class GumbelCDF(Bijector):
|
|||
scale (float, list, numpy.ndarray, Tensor): The scale. Default: 1.0.
|
||||
name (str): The name of the Bijector. Default: 'GumbelCDF'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- **forward**
|
||||
- **inverse**
|
||||
- **forward_log_jacobian**
|
||||
- **backward_log_jacobian**
|
||||
|
||||
It should be notice that the input should be always a tensor,
|
||||
with a shape that can be broadcasted to that of `loc` and `scale`.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Bijector`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -44,7 +57,7 @@ class GumbelCDF(Bijector):
|
|||
|
||||
Raises:
|
||||
TypeError: When the dtype of `loc` or `scale` is not float,
|
||||
and when the dtype of `loc` and `scale` is not same.
|
||||
or when the dtype of `loc` and `scale` is not same.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
|
|
@ -86,18 +99,28 @@ class GumbelCDF(Bijector):
|
|||
self._scale = self._add_parameter(scale, 'scale')
|
||||
check_greater_zero(self._scale, "scale")
|
||||
|
||||
|
||||
self.cast = P.Cast()
|
||||
self.exp = exp_generic
|
||||
self.log = log_generic
|
||||
|
||||
|
||||
@property
|
||||
def loc(self):
|
||||
"""
|
||||
Return the loc parameter of the bijector.
|
||||
|
||||
Output:
|
||||
Tensor, the loc parameter of the bijector.
|
||||
"""
|
||||
return self._loc
|
||||
|
||||
@property
|
||||
def scale(self):
|
||||
"""
|
||||
Return the scale parameter of the bijector.
|
||||
|
||||
Output:
|
||||
Tensor, the scale parameter of the bijector.
|
||||
"""
|
||||
return self._scale
|
||||
|
||||
def extend_repr(self):
|
||||
|
|
|
|||
|
|
@ -26,6 +26,18 @@ class Invert(Bijector):
|
|||
name (str): The name of the Bijector. Default: "". When name is set to "", it is actually
|
||||
'Invert' + bijector.name.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- **forward**
|
||||
- **inverse**
|
||||
- **forward_log_jacobian**
|
||||
- **backward_log_jacobian**
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Bijector`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -73,36 +85,52 @@ class Invert(Bijector):
|
|||
|
||||
def inverse(self, y):
|
||||
"""
|
||||
Forward transformation: transform the input value to another distribution.
|
||||
Perform the inverse transformation of the inverse bijector,
|
||||
namely the forward transformation of the underlying bijector.
|
||||
|
||||
Args:
|
||||
y (Tensor): Tensor of any shape.
|
||||
y (Tensor): the value of the transformed random variable.
|
||||
|
||||
Output:
|
||||
Tensor, the value of the input random variable.
|
||||
"""
|
||||
return self.bijector("forward", y)
|
||||
|
||||
def forward(self, x):
|
||||
"""
|
||||
Inverse transformation: transform the input value back to the original distribution.
|
||||
Perform the forward transformation of the inverse bijector,
|
||||
namely the inverse transformation of the underlying bijector.
|
||||
|
||||
Args:
|
||||
x (Tensor): Tensor of any shape.
|
||||
x (Tensor): the value of the input random variable.
|
||||
|
||||
Output:
|
||||
Tensor, the value of the transformed random variable.
|
||||
"""
|
||||
return self.bijector("inverse", x)
|
||||
|
||||
def inverse_log_jacobian(self, y):
|
||||
"""
|
||||
Logarithm of the derivative of the forward transformation.
|
||||
Logarithm of the derivative of the inverse transformation of the inverse bijector,
|
||||
namely logarithm of the derivative of the forward transformation of the underlying bijector.
|
||||
|
||||
Args:
|
||||
y (Tensor): Tensor of any shape.
|
||||
y (Tensor): the value of the transformed random variable.
|
||||
|
||||
Output:
|
||||
Tensor, logarithm of the derivative of the inverse transformation of the inverse bijector.
|
||||
"""
|
||||
return self.bijector("forward_log_jacobian", y)
|
||||
|
||||
def forward_log_jacobian(self, x):
|
||||
"""
|
||||
Logarithm of the derivative of the inverse transformation.
|
||||
Logarithm of the derivative of the forward transformation of the inverse bijector,
|
||||
namely logarithm of the derivative of the inverse transformation of the underlying bijector.
|
||||
|
||||
Args:
|
||||
x (Tensor): Tensor of any shape.
|
||||
x (Tensor): the value of the input random variable.
|
||||
|
||||
Output:
|
||||
Tensor, logarithm of the derivative of the forward transformation of the inverse bijector.
|
||||
"""
|
||||
return self.bijector("inverse_log_jacobian", x)
|
||||
|
|
|
|||
|
|
@ -37,6 +37,19 @@ class PowerTransform(Bijector):
|
|||
power (float, list, numpy.ndarray, Tensor): The scale factor. Default: 0.
|
||||
name (str): The name of the bijector. Default: 'PowerTransform'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- **forward**
|
||||
- **inverse**
|
||||
- **forward_log_jacobian**
|
||||
- **backward_log_jacobian**
|
||||
|
||||
It should be notice that the input should be always a tensor,
|
||||
with a shape that can be broadcasted to that of `power`.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Bijector`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -92,6 +105,12 @@ class PowerTransform(Bijector):
|
|||
|
||||
@property
|
||||
def power(self):
|
||||
"""
|
||||
Return the power parameter of the bijector.
|
||||
|
||||
Output:
|
||||
Tensor, the power parameter of the bijector.
|
||||
"""
|
||||
return self._power
|
||||
|
||||
def extend_repr(self):
|
||||
|
|
@ -110,7 +129,8 @@ class PowerTransform(Bijector):
|
|||
power_local = self.cast_param_by_value(x, self.power)
|
||||
|
||||
# broad cast the value of x and power
|
||||
ones = self.fill(self.dtypeop(power_local), self.shape(x + power_local), 1.)
|
||||
ones = self.fill(self.dtypeop(power_local),
|
||||
self.shape(x + power_local), 1.)
|
||||
power_local = power_local * ones
|
||||
x = x * ones
|
||||
safe_power = self.select_base(self.equal_base(power_local, 0.),
|
||||
|
|
@ -130,7 +150,8 @@ class PowerTransform(Bijector):
|
|||
power_local = self.cast_param_by_value(y, self.power)
|
||||
|
||||
# broad cast the value of x and power
|
||||
ones = self.fill(self.dtypeop(power_local), self.shape(y + power_local), 1.)
|
||||
ones = self.fill(self.dtypeop(power_local),
|
||||
self.shape(y + power_local), 1.)
|
||||
power_local = power_local * ones
|
||||
y = y * ones
|
||||
safe_power = self.select_base(self.equal_base(power_local, 0.),
|
||||
|
|
@ -159,7 +180,8 @@ class PowerTransform(Bijector):
|
|||
power_local = self.cast_param_by_value(x, self.power)
|
||||
|
||||
# broad cast the value of x and power
|
||||
ones = self.fill(self.dtypeop(power_local), self.shape(x + power_local), 1.)
|
||||
ones = self.fill(self.dtypeop(power_local),
|
||||
self.shape(x + power_local), 1.)
|
||||
power_local = power_local * ones
|
||||
x = x * ones
|
||||
|
||||
|
|
|
|||
|
|
@ -33,6 +33,19 @@ class ScalarAffine(Bijector):
|
|||
shift (float, list, numpy.ndarray, Tensor): The shift factor. Default: 0.0.
|
||||
name (str): The name of the bijector. Default: 'ScalarAffine'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- **forward**
|
||||
- **inverse**
|
||||
- **forward_log_jacobian**
|
||||
- **backward_log_jacobian**
|
||||
|
||||
It should be notice that the input should be always a tensor,
|
||||
with a shape that can be broadcasted to that of `shift` and `scale`.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Bijector`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -94,10 +107,22 @@ class ScalarAffine(Bijector):
|
|||
|
||||
@property
|
||||
def scale(self):
|
||||
"""
|
||||
Return the scale parameter of the bijector.
|
||||
|
||||
Output:
|
||||
Tensor, the scale parameter of the bijector.
|
||||
"""
|
||||
return self._scale
|
||||
|
||||
@property
|
||||
def shift(self):
|
||||
"""
|
||||
Return the shift parameter of the bijector.
|
||||
|
||||
Output:
|
||||
Tensor, the shift parameter of the bijector.
|
||||
"""
|
||||
return self._shift
|
||||
|
||||
def extend_repr(self):
|
||||
|
|
|
|||
|
|
@ -34,6 +34,19 @@ class Softplus(Bijector):
|
|||
sharpness (float, list, numpy.ndarray, Tensor): The scale factor. Default: 1.0.
|
||||
name (str): The name of the Bijector. Default: 'Softplus'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- **forward**
|
||||
- **inverse**
|
||||
- **forward_log_jacobian**
|
||||
- **backward_log_jacobian**
|
||||
|
||||
It should be notice that the input should be always a tensor,
|
||||
with a shape that can be broadcasted to that of `sharpness`.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Bijector`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -127,6 +140,12 @@ class Softplus(Bijector):
|
|||
|
||||
@property
|
||||
def sharpness(self):
|
||||
"""
|
||||
Return the sharpness parameter of the bijector.
|
||||
|
||||
Output:
|
||||
Tensor, the sharpness parameter of the bijector.
|
||||
"""
|
||||
return self._sharpness
|
||||
|
||||
def extend_repr(self):
|
||||
|
|
|
|||
|
|
@ -25,6 +25,8 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
class Bernoulli(Distribution):
|
||||
"""
|
||||
Bernoulli Distribution.
|
||||
A Bernoulli Distribution is a discrete distribution with the range {0, 1}
|
||||
and the probability mass function as :math:`P(X = 0) = p, P(X = 1) = 1-p`.
|
||||
|
||||
Args:
|
||||
probs (float, list, numpy.ndarray, Tensor): The probability of that the outcome is 1. Default: None.
|
||||
|
|
@ -32,6 +34,18 @@ class Bernoulli(Distribution):
|
|||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.int32.
|
||||
name (str): The name of the distribution. Default: 'Bernoulli'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`
|
||||
- `mean`, `sd`, `var`, and `entropy`
|
||||
- `kl_loss` and `cross_entropy`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -39,6 +53,10 @@ class Bernoulli(Distribution):
|
|||
`probs` must be a proper probability (0 < p < 1).
|
||||
`dist_spec_args` is `probs`.
|
||||
|
||||
Raises:
|
||||
ValueError: When p <= 0 or p >=1.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -91,7 +109,7 @@ class Bernoulli(Distribution):
|
|||
>>> # dist (str): the name of the distribution. Only 'Bernoulli' is supported.
|
||||
>>> # probs1_b (Tensor): the probability of success of distribution b.
|
||||
>>> # probs1_a (Tensor): the probability of success of distribution a. Default: self.probs.
|
||||
>>> # Examples of kl_loss. `cross_entropy` is similar.
|
||||
>>> # Examples of `kl_loss`. `cross_entropy` is similar.
|
||||
>>> ans = b1.kl_loss('Bernoulli', probs_b)
|
||||
>>> print(ans.shape)
|
||||
(3,)
|
||||
|
|
@ -165,6 +183,9 @@ class Bernoulli(Distribution):
|
|||
"""
|
||||
Return the probability of that the outcome is 1
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the probs of the distribution.
|
||||
"""
|
||||
return self._probs
|
||||
|
||||
|
|
|
|||
|
|
@ -25,18 +25,36 @@ from ._utils.custom_ops import log_generic
|
|||
|
||||
|
||||
class Beta(Distribution):
|
||||
"""
|
||||
r"""
|
||||
Beta distribution.
|
||||
A Beta distributio is a continuous distribution with the range :math:`[0, 1]` and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, \alpha, \beta) = x^\alpha (1-x)^{\beta - 1} / B(\alpha, \beta),
|
||||
|
||||
where :math:`B` is the Beta function.
|
||||
|
||||
Args:
|
||||
concentration1 (list, numpy.ndarray, Tensor): The concentration1,
|
||||
also know as alpha of the Beta distribution.
|
||||
concentration0 (list, numpy.ndarray, Tensor): The concentration0, also know as
|
||||
beta of the Beta distribution.
|
||||
concentration1 (int, float, list, numpy.ndarray, Tensor): The concentration1,
|
||||
also know as alpha of the Beta distribution. Default: None.
|
||||
concentration0 (int, float, list, numpy.ndarray, Tensor): The concentration0, also know as
|
||||
beta of the Beta distribution. Default: None.
|
||||
seed (int): The seed used in sampling. The global seed is used if it is None. Default: None.
|
||||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Beta'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob` and `log_prob`
|
||||
- `mean`, `sd`, `var`, and `entropy`
|
||||
- `kl_loss` and `cross_entropy`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend``
|
||||
|
||||
|
|
@ -45,6 +63,10 @@ class Beta(Distribution):
|
|||
`dist_spec_args` are `concentration1` and `concentration0`.
|
||||
`dtype` must be a float type because Beta distributions are continuous.
|
||||
|
||||
Raises:
|
||||
ValueError: When concentration1 <= 0 or concentration0 >=1.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -145,10 +167,12 @@ class Beta(Distribution):
|
|||
Constructor of Beta.
|
||||
"""
|
||||
param = dict(locals())
|
||||
param['param_dict'] = {'concentration1': concentration1, 'concentration0': concentration0}
|
||||
param['param_dict'] = {
|
||||
'concentration1': concentration1, 'concentration0': concentration0}
|
||||
|
||||
valid_dtype = mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
|
||||
# As some operators can't accept scalar input, check the type here
|
||||
if isinstance(concentration0, float):
|
||||
|
|
@ -158,8 +182,10 @@ class Beta(Distribution):
|
|||
|
||||
super(Beta, self).__init__(seed, dtype, name, param)
|
||||
|
||||
self._concentration1 = self._add_parameter(concentration1, 'concentration1')
|
||||
self._concentration0 = self._add_parameter(concentration0, 'concentration0')
|
||||
self._concentration1 = self._add_parameter(
|
||||
concentration1, 'concentration1')
|
||||
self._concentration0 = self._add_parameter(
|
||||
concentration0, 'concentration0')
|
||||
if self._concentration1 is not None:
|
||||
check_greater_zero(self._concentration1, "concentration1")
|
||||
if self._concentration0 is not None:
|
||||
|
|
@ -183,7 +209,8 @@ class Beta(Distribution):
|
|||
def extend_repr(self):
|
||||
"""Display instance object as string."""
|
||||
if self.is_scalar_batch:
|
||||
s = 'concentration1 = {}, concentration0 = {}'.format(self._concentration1, self._concentration0)
|
||||
s = 'concentration1 = {}, concentration0 = {}'.format(
|
||||
self._concentration1, self._concentration0)
|
||||
else:
|
||||
s = 'batch_shape = {}'.format(self._broadcast_shape)
|
||||
return s
|
||||
|
|
@ -193,6 +220,9 @@ class Beta(Distribution):
|
|||
"""
|
||||
Return the concentration1, also know as the alpha of the Beta distribution,
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the concentration1 parameter of the distribution.
|
||||
"""
|
||||
return self._concentration1
|
||||
|
||||
|
|
@ -201,6 +231,9 @@ class Beta(Distribution):
|
|||
"""
|
||||
Return the concentration0, also know as the beta of the Beta distribution,
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the concentration2 parameter of the distribution.
|
||||
"""
|
||||
return self._concentration0
|
||||
|
||||
|
|
@ -222,14 +255,16 @@ class Beta(Distribution):
|
|||
"""
|
||||
The mean of the distribution.
|
||||
"""
|
||||
concentration1, concentration0 = self._check_param_type(concentration1, concentration0)
|
||||
concentration1, concentration0 = self._check_param_type(
|
||||
concentration1, concentration0)
|
||||
return concentration1 / (concentration1 + concentration0)
|
||||
|
||||
def _var(self, concentration1=None, concentration0=None):
|
||||
"""
|
||||
The variance of the distribution.
|
||||
"""
|
||||
concentration1, concentration0 = self._check_param_type(concentration1, concentration0)
|
||||
concentration1, concentration0 = self._check_param_type(
|
||||
concentration1, concentration0)
|
||||
total_concentration = concentration1 + concentration0
|
||||
return concentration1 * concentration0 / (self.pow(total_concentration, 2) * (total_concentration + 1.))
|
||||
|
||||
|
|
@ -237,7 +272,8 @@ class Beta(Distribution):
|
|||
"""
|
||||
The mode of the distribution.
|
||||
"""
|
||||
concentration1, concentration0 = self._check_param_type(concentration1, concentration0)
|
||||
concentration1, concentration0 = self._check_param_type(
|
||||
concentration1, concentration0)
|
||||
comp1 = self.greater(concentration1, 1.)
|
||||
comp2 = self.greater(concentration0, 1.)
|
||||
cond = self.logicaland(comp1, comp2)
|
||||
|
|
@ -254,12 +290,13 @@ class Beta(Distribution):
|
|||
H(X) = \log(\Beta(\alpha, \beta)) - (\alpha - 1) * \digamma(\alpha)
|
||||
- (\beta - 1) * \digamma(\beta) + (\alpha + \beta - 2) * \digamma(\alpha + \beta)
|
||||
"""
|
||||
concentration1, concentration0 = self._check_param_type(concentration1, concentration0)
|
||||
concentration1, concentration0 = self._check_param_type(
|
||||
concentration1, concentration0)
|
||||
total_concentration = concentration1 + concentration0
|
||||
return self.lbeta(concentration1, concentration0) \
|
||||
- (concentration1 - 1.) * self.digamma(concentration1) \
|
||||
- (concentration0 - 1.) * self.digamma(concentration0) \
|
||||
+ (total_concentration - 2.) * self.digamma(total_concentration)
|
||||
- (concentration1 - 1.) * self.digamma(concentration1) \
|
||||
- (concentration0 - 1.) * self.digamma(concentration0) \
|
||||
+ (total_concentration - 2.) * self.digamma(total_concentration)
|
||||
|
||||
def _cross_entropy(self, dist, concentration1_b, concentration0_b, concentration1_a=None, concentration0_a=None):
|
||||
r"""
|
||||
|
|
@ -274,7 +311,8 @@ class Beta(Distribution):
|
|||
"""
|
||||
check_distribution_name(dist, 'Beta')
|
||||
return self._entropy(concentration1_a, concentration0_a) \
|
||||
+ self._kl_loss(dist, concentration1_b, concentration0_b, concentration1_a, concentration0_a)
|
||||
+ self._kl_loss(dist, concentration1_b, concentration0_b,
|
||||
concentration1_a, concentration0_a)
|
||||
|
||||
def _log_prob(self, value, concentration1=None, concentration0=None):
|
||||
r"""
|
||||
|
|
@ -290,9 +328,10 @@ class Beta(Distribution):
|
|||
"""
|
||||
value = self._check_value(value, 'value')
|
||||
value = self.cast(value, self.dtype)
|
||||
concentration1, concentration0 = self._check_param_type(concentration1, concentration0)
|
||||
concentration1, concentration0 = self._check_param_type(
|
||||
concentration1, concentration0)
|
||||
log_unnormalized_prob = (concentration1 - 1.) * self.log(value) \
|
||||
+ (concentration0 - 1.) * self.log1p(self.neg(value))
|
||||
+ (concentration0 - 1.) * self.log1p(self.neg(value))
|
||||
return log_unnormalized_prob - self.lbeta(concentration1, concentration0)
|
||||
|
||||
def _kl_loss(self, dist, concentration1_b, concentration0_b, concentration1_a=None, concentration0_a=None):
|
||||
|
|
@ -313,19 +352,23 @@ class Beta(Distribution):
|
|||
+ \digamma(\alpha_{a} + \beta_{a}) * (\alpha_{b} + \beta_{b} - \alpha_{a} - \beta_{a})
|
||||
"""
|
||||
check_distribution_name(dist, 'Beta')
|
||||
concentration1_b = self._check_value(concentration1_b, 'concentration1_b')
|
||||
concentration0_b = self._check_value(concentration0_b, 'concentration0_b')
|
||||
concentration1_b = self._check_value(
|
||||
concentration1_b, 'concentration1_b')
|
||||
concentration0_b = self._check_value(
|
||||
concentration0_b, 'concentration0_b')
|
||||
concentration1_b = self.cast(concentration1_b, self.parameter_type)
|
||||
concentration0_b = self.cast(concentration0_b, self.parameter_type)
|
||||
concentration1_a, concentration0_a = self._check_param_type(concentration1_a, concentration0_a)
|
||||
concentration1_a, concentration0_a = self._check_param_type(
|
||||
concentration1_a, concentration0_a)
|
||||
total_concentration_a = concentration1_a + concentration0_a
|
||||
total_concentration_b = concentration1_b + concentration0_b
|
||||
log_normalization_a = self.lbeta(concentration1_a, concentration0_a)
|
||||
log_normalization_b = self.lbeta(concentration1_b, concentration0_b)
|
||||
return (log_normalization_b - log_normalization_a) \
|
||||
- (self.digamma(concentration1_a) * (concentration1_b - concentration1_a)) \
|
||||
- (self.digamma(concentration0_a) * (concentration0_b - concentration0_a)) \
|
||||
+ (self.digamma(total_concentration_a) * (total_concentration_b - total_concentration_a))
|
||||
- (self.digamma(concentration1_a) * (concentration1_b - concentration1_a)) \
|
||||
- (self.digamma(concentration0_a) * (concentration0_b - concentration0_a)) \
|
||||
+ (self.digamma(total_concentration_a) *
|
||||
(total_concentration_b - total_concentration_a))
|
||||
|
||||
def _sample(self, shape=(), concentration1=None, concentration0=None):
|
||||
"""
|
||||
|
|
@ -340,7 +383,8 @@ class Beta(Distribution):
|
|||
Tensor, with the shape being shape + batch_shape.
|
||||
"""
|
||||
shape = self.checktuple(shape, 'shape')
|
||||
concentration1, concentration0 = self._check_param_type(concentration1, concentration0)
|
||||
concentration1, concentration0 = self._check_param_type(
|
||||
concentration1, concentration0)
|
||||
batch_shape = self.shape(concentration1 + concentration0)
|
||||
origin_shape = shape + batch_shape
|
||||
if origin_shape == ():
|
||||
|
|
@ -348,8 +392,10 @@ class Beta(Distribution):
|
|||
else:
|
||||
sample_shape = origin_shape
|
||||
ones = self.fill(self.dtype, sample_shape, 1.0)
|
||||
sample_gamma1 = C.gamma(sample_shape, alpha=concentration1, beta=ones, seed=self.seed)
|
||||
sample_gamma2 = C.gamma(sample_shape, alpha=concentration0, beta=ones, seed=self.seed)
|
||||
sample_gamma1 = C.gamma(
|
||||
sample_shape, alpha=concentration1, beta=ones, seed=self.seed)
|
||||
sample_gamma2 = C.gamma(
|
||||
sample_shape, alpha=concentration0, beta=ones, seed=self.seed)
|
||||
sample_beta = sample_gamma1 / (sample_gamma1 + sample_gamma2)
|
||||
value = self.cast(sample_beta, self.dtype)
|
||||
if origin_shape == ():
|
||||
|
|
|
|||
|
|
@ -29,20 +29,37 @@ from ._utils.custom_ops import exp_generic, log_generic, broadcast_to
|
|||
|
||||
class Categorical(Distribution):
|
||||
"""
|
||||
Create a categorical distribution parameterized by event probabilities.
|
||||
Categorical distribution.
|
||||
A Categorical Distribution is a discrete distribution with the range {1, 2, ..., k}
|
||||
and the probability mass function as :math:`P(X = i) = p_i, i = 1, ..., k`.
|
||||
|
||||
Args:
|
||||
probs (Tensor, list, numpy.ndarray): Event probabilities.
|
||||
probs (Tensor, list, numpy.ndarray): Event probabilities. Default: None.
|
||||
seed (int): The global seed is used in sampling. Global seed is used if it is None. Default: None.
|
||||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.int32.
|
||||
name (str): The name of the distribution. Default: Categorical.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`
|
||||
- `mean`, `sd`, `var`, and `entropy`
|
||||
- `kl_loss` and `cross_entropy`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
Note:
|
||||
`probs` must have rank at least 1, values are proper probabilities and sum to 1.
|
||||
|
||||
Raises:
|
||||
ValueError: When the sum of all elements in `probs` is not 1.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -95,7 +112,7 @@ class Categorical(Distribution):
|
|||
>>> # dist (str): the name of the distribution. Only 'Categorical' is supported.
|
||||
>>> # probs_b (Tensor): event probabilities of distribution b.
|
||||
>>> # probs (Tensor): event probabilities of distribution a. Default: self.probs.
|
||||
>>> # Examples of kl_loss. `cross_entropy` is similar.
|
||||
>>> # Examples of `kl_loss`, `cross_entropy` is similar.
|
||||
>>> ans = ca1.kl_loss('Categorical', probs_b)
|
||||
>>> print(ans.shape)
|
||||
()
|
||||
|
|
@ -172,6 +189,9 @@ class Categorical(Distribution):
|
|||
def probs(self):
|
||||
"""
|
||||
Return the probability after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the probs of the distribution.
|
||||
"""
|
||||
return self._probs
|
||||
|
||||
|
|
|
|||
|
|
@ -24,8 +24,15 @@ from ._utils.custom_ops import exp_generic, log_generic, log1p_generic
|
|||
|
||||
|
||||
class Cauchy(Distribution):
|
||||
"""
|
||||
r"""
|
||||
Cauchy distribution.
|
||||
A Cauchy distributio is a continuous distribution with the range :math:`[0, 1]`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / \pi b(1 - ((x - a)/b)^2),
|
||||
|
||||
where a and b are loc and scale parameter respectively.
|
||||
|
||||
Args:
|
||||
loc (int, float, list, numpy.ndarray, Tensor): The location of the Cauchy distribution.
|
||||
|
|
@ -34,6 +41,18 @@ class Cauchy(Distribution):
|
|||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Cauchy'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`
|
||||
- `mode` and `entropy`
|
||||
- `kl_loss` and `cross_entropy`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend``
|
||||
|
||||
|
|
@ -43,6 +62,10 @@ class Cauchy(Distribution):
|
|||
`dtype` must be a float type because Cauchy distributions are continuous.
|
||||
Cauchy distribution is not supported on GPU backend.
|
||||
|
||||
Raises:
|
||||
ValueError: When scale <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -144,7 +167,8 @@ class Cauchy(Distribution):
|
|||
param = dict(locals())
|
||||
param['param_dict'] = {'loc': loc, 'scale': scale}
|
||||
valid_dtype = mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
super(Cauchy, self).__init__(seed, dtype, name, param)
|
||||
|
||||
self._loc = self._add_parameter(loc, 'loc')
|
||||
|
|
@ -171,11 +195,11 @@ class Cauchy(Distribution):
|
|||
|
||||
self.entropy_const = np.log(4 * np.pi)
|
||||
|
||||
|
||||
def extend_repr(self):
|
||||
"""Display instance object as string."""
|
||||
if self.is_scalar_batch:
|
||||
str_info = 'location = {}, scale = {}'.format(self._loc, self._scale)
|
||||
str_info = 'location = {}, scale = {}'.format(
|
||||
self._loc, self._scale)
|
||||
else:
|
||||
str_info = 'batch_shape = {}'.format(self._broadcast_shape)
|
||||
return str_info
|
||||
|
|
@ -184,6 +208,9 @@ class Cauchy(Distribution):
|
|||
def loc(self):
|
||||
"""
|
||||
Return the location of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the loc parameter of the distribution.
|
||||
"""
|
||||
return self._loc
|
||||
|
||||
|
|
@ -191,6 +218,9 @@ class Cauchy(Distribution):
|
|||
def scale(self):
|
||||
"""
|
||||
Return the scale of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the scale parameter of the distribution.
|
||||
"""
|
||||
return self._scale
|
||||
|
||||
|
|
@ -320,7 +350,7 @@ class Cauchy(Distribution):
|
|||
sum_square = self.sq(scale_a + scale_b)
|
||||
square_diff = self.sq(loc_a - loc_b)
|
||||
return self.log(sum_square + square_diff) - \
|
||||
self.log(self.const(4.0)) - self.log(scale_a) - self.log(scale_b)
|
||||
self.log(self.const(4.0)) - self.log(scale_a) - self.log(scale_b)
|
||||
|
||||
def _cross_entropy(self, dist, loc_b, scale_b, loc_a=None, scale_a=None):
|
||||
r"""
|
||||
|
|
|
|||
|
|
@ -437,7 +437,7 @@ class Distribution(Cell):
|
|||
|
||||
def cdf(self, value, *args, **kwargs):
|
||||
"""
|
||||
Evaluate the cdf at given value.
|
||||
Evaluate the cumulative distribution function(cdf) at given value.
|
||||
|
||||
Args:
|
||||
value (Tensor): value to be evaluated.
|
||||
|
|
@ -447,6 +447,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its dist_spec_args through
|
||||
`args` or `kwargs`.
|
||||
|
||||
Output:
|
||||
Tensor, the cdf of the distribution.
|
||||
"""
|
||||
return self._call_cdf(value, *args, **kwargs)
|
||||
|
||||
|
|
@ -479,7 +482,7 @@ class Distribution(Cell):
|
|||
|
||||
def log_cdf(self, value, *args, **kwargs):
|
||||
"""
|
||||
Evaluate the log cdf at given value.
|
||||
Evaluate the log the cumulative distribution function(cdf) at given value.
|
||||
|
||||
Args:
|
||||
value (Tensor): value to be evaluated.
|
||||
|
|
@ -489,6 +492,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its dist_spec_args through
|
||||
`args` or `kwargs`.
|
||||
|
||||
Output:
|
||||
Tensor, the log cdf of the distribution.
|
||||
"""
|
||||
return self._call_log_cdf(value, *args, **kwargs)
|
||||
|
||||
|
|
@ -513,6 +519,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its dist_spec_args through
|
||||
`args` or `kwargs`.
|
||||
|
||||
Output:
|
||||
Tensor, the survival function of the distribution.
|
||||
"""
|
||||
return self._call_survival(value, *args, **kwargs)
|
||||
|
||||
|
|
@ -546,6 +555,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its dist_spec_args through
|
||||
`args` or `kwargs`.
|
||||
|
||||
Output:
|
||||
Tensor, the log survival function of the distribution.
|
||||
"""
|
||||
return self._call_log_survival(value, *args, **kwargs)
|
||||
|
||||
|
|
@ -573,6 +585,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
dist_spec_args of distribution b must be passed to the function through `args` or `kwargs`.
|
||||
Passing in dist_spec_args of distribution a is optional.
|
||||
|
||||
Output:
|
||||
Tensor, the kl loss function of the distribution.
|
||||
"""
|
||||
return self._kl_loss(dist, *args, **kwargs)
|
||||
|
||||
|
|
@ -590,6 +605,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its *dist_spec_args* through
|
||||
*args* or *kwargs*.
|
||||
|
||||
Output:
|
||||
Tensor, the mean of the distribution.
|
||||
"""
|
||||
return self._mean(*args, **kwargs)
|
||||
|
||||
|
|
@ -607,6 +625,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its *dist_spec_args* through
|
||||
*args* or *kwargs*.
|
||||
|
||||
Output:
|
||||
Tensor, the mode of the distribution.
|
||||
"""
|
||||
return self._mode(*args, **kwargs)
|
||||
|
||||
|
|
@ -616,11 +637,14 @@ class Distribution(Cell):
|
|||
|
||||
Args:
|
||||
*args (list): the list of positional arguments forwarded to subclasses.
|
||||
**kwargs (dict: the dictionary of keyword arguments forwarded to subclasses.
|
||||
**kwargs (dict): the dictionary of keyword arguments forwarded to subclasses.
|
||||
|
||||
Note:
|
||||
A distribution can be optionally passed to the function by passing its *dist_spec_args* through
|
||||
*args* or *kwargs*.
|
||||
|
||||
Output:
|
||||
Tensor, the standard deviation of the distribution.
|
||||
"""
|
||||
return self._call_sd(*args, **kwargs)
|
||||
|
||||
|
|
@ -635,6 +659,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its *dist_spec_args* through
|
||||
*args* or *kwargs*.
|
||||
|
||||
Output:
|
||||
Tensor, the variance of the distribution.
|
||||
"""
|
||||
return self._call_var(*args, **kwargs)
|
||||
|
||||
|
|
@ -670,6 +697,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its *dist_spec_args* through
|
||||
*args* or *kwargs*.
|
||||
|
||||
Output:
|
||||
Tensor, the entropy of the distribution.
|
||||
"""
|
||||
return self._entropy(*args, **kwargs)
|
||||
|
||||
|
|
@ -685,6 +715,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
dist_spec_args of distribution b must be passed to the function through `args` or `kwargs`.
|
||||
Passing in dist_spec_args of distribution a is optional.
|
||||
|
||||
Output:
|
||||
Tensor, the cross_entropy of two distributions.
|
||||
"""
|
||||
return self._call_cross_entropy(dist, *args, **kwargs)
|
||||
|
||||
|
|
@ -712,6 +745,9 @@ class Distribution(Cell):
|
|||
Note:
|
||||
A distribution can be optionally passed to the function by passing its *dist_spec_args* through
|
||||
*args* or *kwargs*.
|
||||
|
||||
Output:
|
||||
Tensor, the sample generated from the distribution.
|
||||
"""
|
||||
return self._sample(*args, **kwargs)
|
||||
|
||||
|
|
|
|||
|
|
@ -24,15 +24,34 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class Exponential(Distribution):
|
||||
"""
|
||||
Example class: Exponential Distribution.
|
||||
r"""
|
||||
Exponential Distribution.
|
||||
An Exponential distributio is a continuous distribution with the range :math:`[0, 1]`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, \lambda) = \lambda \exp(-\lambda x),
|
||||
|
||||
where :math:`\lambda` is the rate of the distribution.
|
||||
|
||||
Args:
|
||||
rate (float, list, numpy.ndarray, Tensor): The inverse scale. Default: None.
|
||||
rate (int, float, list, numpy.ndarray, Tensor): The inverse scale. Default: None.
|
||||
seed (int): The seed used in sampling. The global seed is used if it is None. Default: None.
|
||||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Exponential'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`
|
||||
- `mean`, `sd`, `var`, and `entropy`
|
||||
- `kl_loss` and `cross_entropy`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -41,6 +60,10 @@ class Exponential(Distribution):
|
|||
`dist_spec_args` is `rate`.
|
||||
`dtype` must be a float type because Exponential distributions are continuous.
|
||||
|
||||
Raises:
|
||||
ValueError: When rate <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -133,7 +156,8 @@ class Exponential(Distribution):
|
|||
param = dict(locals())
|
||||
param['param_dict'] = {'rate': rate}
|
||||
valid_dtype = mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
super(Exponential, self).__init__(seed, dtype, name, param)
|
||||
|
||||
self._rate = self._add_parameter(rate, 'rate')
|
||||
|
|
@ -167,6 +191,9 @@ class Exponential(Distribution):
|
|||
def rate(self):
|
||||
"""
|
||||
Return `rate` of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the rate parameter of the distribution.
|
||||
"""
|
||||
return self._rate
|
||||
|
||||
|
|
|
|||
|
|
@ -25,18 +25,38 @@ from ._utils.custom_ops import log_generic
|
|||
|
||||
|
||||
class Gamma(Distribution):
|
||||
"""
|
||||
r"""
|
||||
Gamma distribution.
|
||||
A Gamma distributio is a continuous distribution with the range :math:`[0, 1]`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, \alpha, \beta) = \beta^\alpha / \Gamma(\alpha) x^{\alpha - 1} \exp(-\beta x).
|
||||
|
||||
where :math:`G` is the Gamma function,
|
||||
and :math:`\alpha, \beta` are the concentration and the rate of the distribution respectively.
|
||||
|
||||
Args:
|
||||
concentration (list, numpy.ndarray, Tensor): The concentration,
|
||||
concentration (int, float, list, numpy.ndarray, Tensor): The concentration,
|
||||
also know as alpha of the Gamma distribution. Default: None.
|
||||
rate (list, numpy.ndarray, Tensor): The rate, also know as
|
||||
rate (int, float, list, numpy.ndarray, Tensor): The rate, also know as
|
||||
beta of the Gamma distribution. Default: None.
|
||||
seed (int): The seed used in sampling. The global seed is used if it is None. Default: None.
|
||||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Gamma'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is 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`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend``
|
||||
|
||||
|
|
@ -45,6 +65,10 @@ class Gamma(Distribution):
|
|||
`dist_spec_args` are `concentration` and `rate`.
|
||||
`dtype` must be a float type because Gamma distributions are continuous.
|
||||
|
||||
Raises:
|
||||
ValueError: When concentration <= 0 or rate <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -147,7 +171,8 @@ class Gamma(Distribution):
|
|||
param = dict(locals())
|
||||
param['param_dict'] = {'concentration': concentration, 'rate': rate}
|
||||
valid_dtype = mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
|
||||
# As some operators can't accept scalar input, check the type here
|
||||
if isinstance(concentration, (int, float)):
|
||||
|
|
@ -157,7 +182,8 @@ class Gamma(Distribution):
|
|||
|
||||
super(Gamma, self).__init__(seed, dtype, name, param)
|
||||
|
||||
self._concentration = self._add_parameter(concentration, 'concentration')
|
||||
self._concentration = self._add_parameter(
|
||||
concentration, 'concentration')
|
||||
self._rate = self._add_parameter(rate, 'rate')
|
||||
if self._concentration is not None:
|
||||
check_greater_zero(self._concentration, "concentration")
|
||||
|
|
@ -182,7 +208,8 @@ class Gamma(Distribution):
|
|||
def extend_repr(self):
|
||||
"""Display instance object as string."""
|
||||
if self.is_scalar_batch:
|
||||
s = 'concentration = {}, rate = {}'.format(self._concentration, self._rate)
|
||||
s = 'concentration = {}, rate = {}'.format(
|
||||
self._concentration, self._rate)
|
||||
else:
|
||||
s = 'batch_shape = {}'.format(self._broadcast_shape)
|
||||
return s
|
||||
|
|
@ -192,6 +219,9 @@ class Gamma(Distribution):
|
|||
"""
|
||||
Return the concentration, also know as the alpha of the Gamma distribution,
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the concentration parameter of the distribution.
|
||||
"""
|
||||
return self._concentration
|
||||
|
||||
|
|
@ -200,6 +230,9 @@ class Gamma(Distribution):
|
|||
"""
|
||||
Return the rate, also know as the beta of the Gamma distribution,
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the rate parameter of the distribution.
|
||||
"""
|
||||
return self._rate
|
||||
|
||||
|
|
@ -244,7 +277,8 @@ class Gamma(Distribution):
|
|||
"""
|
||||
concentration, rate = self._check_param_type(concentration, rate)
|
||||
mode = (concentration - 1.) / rate
|
||||
nan = self.fill(self.dtypeop(concentration), self.shape(concentration), np.nan)
|
||||
nan = self.fill(self.dtypeop(concentration),
|
||||
self.shape(concentration), np.nan)
|
||||
comp = self.greater(concentration, 1.)
|
||||
return self.select(comp, mode, nan)
|
||||
|
||||
|
|
@ -257,7 +291,7 @@ class Gamma(Distribution):
|
|||
"""
|
||||
concentration, rate = self._check_param_type(concentration, rate)
|
||||
return concentration - self.log(rate) + self.lgamma(concentration) \
|
||||
+ (1. - concentration) * self.digamma(concentration)
|
||||
+ (1. - concentration) * self.digamma(concentration)
|
||||
|
||||
def _cross_entropy(self, dist, concentration_b, rate_b, concentration_a=None, rate_a=None):
|
||||
r"""
|
||||
|
|
@ -272,7 +306,8 @@ class Gamma(Distribution):
|
|||
"""
|
||||
check_distribution_name(dist, 'Gamma')
|
||||
return self._entropy(concentration_a, rate_a) +\
|
||||
self._kl_loss(dist, concentration_b, rate_b, concentration_a, rate_a)
|
||||
self._kl_loss(dist, concentration_b, rate_b,
|
||||
concentration_a, rate_a)
|
||||
|
||||
def _log_prob(self, value, concentration=None, rate=None):
|
||||
r"""
|
||||
|
|
@ -289,8 +324,10 @@ class Gamma(Distribution):
|
|||
value = self._check_value(value, 'value')
|
||||
value = self.cast(value, self.dtype)
|
||||
concentration, rate = self._check_param_type(concentration, rate)
|
||||
unnormalized_log_prob = (concentration - 1.) * self.log(value) - rate * value
|
||||
log_normalization = self.lgamma(concentration) - concentration * self.log(rate)
|
||||
unnormalized_log_prob = (concentration - 1.) * \
|
||||
self.log(value) - rate * value
|
||||
log_normalization = self.lgamma(
|
||||
concentration) - concentration * self.log(rate)
|
||||
return unnormalized_log_prob - log_normalization
|
||||
|
||||
def _cdf(self, value, concentration=None, rate=None):
|
||||
|
|
@ -332,11 +369,12 @@ class Gamma(Distribution):
|
|||
rate_b = self._check_value(rate_b, 'rate_b')
|
||||
concentration_b = self.cast(concentration_b, self.parameter_type)
|
||||
rate_b = self.cast(rate_b, self.parameter_type)
|
||||
concentration_a, rate_a = self._check_param_type(concentration_a, rate_a)
|
||||
concentration_a, rate_a = self._check_param_type(
|
||||
concentration_a, rate_a)
|
||||
return (concentration_a - concentration_b) * self.digamma(concentration_a) \
|
||||
+ self.lgamma(concentration_b) - self.lgamma(concentration_a) \
|
||||
+ concentration_b * self.log(rate_a) - concentration_b * self.log(rate_b) \
|
||||
+ concentration_a * (rate_b / rate_a - 1.)
|
||||
+ self.lgamma(concentration_b) - self.lgamma(concentration_a) \
|
||||
+ concentration_b * self.log(rate_a) - concentration_b * self.log(rate_b) \
|
||||
+ concentration_a * (rate_b / rate_a - 1.)
|
||||
|
||||
def _sample(self, shape=(), concentration=None, rate=None):
|
||||
"""
|
||||
|
|
|
|||
|
|
@ -26,16 +26,29 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
class Geometric(Distribution):
|
||||
"""
|
||||
Geometric Distribution.
|
||||
|
||||
A Geometric Distribution is a discrete distribution with the range as the non-negative integers,
|
||||
and the probability mass function as :math:`P(X = i) = p(1-p)^{i-1}, i = 1, 2, ...`.
|
||||
It represents that there are k failures before the first success, namely that there are in total k+1 Bernoulli
|
||||
trials when the first success is achieved.
|
||||
|
||||
Args:
|
||||
probs (float, list, numpy.ndarray, Tensor): The probability of success. Default: None.
|
||||
probs (int, float, list, numpy.ndarray, Tensor): The probability of success. Default: None.
|
||||
seed (int): The seed used in sampling. Global seed is used if it is None. Default: None.
|
||||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.int32.
|
||||
name (str): The name of the distribution. Default: 'Geometric'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is 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`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -43,6 +56,9 @@ class Geometric(Distribution):
|
|||
`probs` must be a proper probability (0 < p < 1).
|
||||
`dist_spec_args` is `probs`.
|
||||
|
||||
Raises:
|
||||
ValueError: When p <= 0 or p >= 1.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -138,7 +154,8 @@ class Geometric(Distribution):
|
|||
param = dict(locals())
|
||||
param['param_dict'] = {'probs': probs}
|
||||
valid_dtype = mstype.int_type + mstype.uint_type + mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
super(Geometric, self).__init__(seed, dtype, name, param)
|
||||
|
||||
self._probs = self._add_parameter(probs, 'probs')
|
||||
|
|
@ -176,6 +193,9 @@ class Geometric(Distribution):
|
|||
def probs(self):
|
||||
"""
|
||||
Return the probability of success of the Bernoulli trial, after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the probs parameter of the distribution.
|
||||
"""
|
||||
return self._probs
|
||||
|
||||
|
|
|
|||
|
|
@ -26,16 +26,31 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class Gumbel(TransformedDistribution):
|
||||
"""
|
||||
r"""
|
||||
Gumbel distribution.
|
||||
A Gumbel distributio is a continuous distribution with the range :math:`[0, 1]`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / b \exp(\exp(-(x - a) / b) - x),
|
||||
|
||||
where a and b are loc and scale parameter respectively.
|
||||
|
||||
Args:
|
||||
loc (float, list, numpy.ndarray, Tensor): The location of Gumbel distribution.
|
||||
scale (float, list, numpy.ndarray, Tensor): The scale of Gumbel distribution.
|
||||
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: None.
|
||||
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 api is 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`
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -45,6 +60,10 @@ class Gumbel(TransformedDistribution):
|
|||
`dtype` must be a float type because Gumbel distributions are continuous.
|
||||
`kl_loss` and `cross_entropy` are not supported on GPU backend.
|
||||
|
||||
Raises:
|
||||
ValueError: When scale <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -72,7 +91,8 @@ class Gumbel(TransformedDistribution):
|
|||
Constructor of Gumbel distribution.
|
||||
"""
|
||||
valid_dtype = mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
gumbel_cdf = msb.GumbelCDF(loc, scale)
|
||||
super(Gumbel, self).__init__(
|
||||
distribution=msd.Uniform(0.0, 1.0, dtype=dtype),
|
||||
|
|
@ -101,6 +121,9 @@ class Gumbel(TransformedDistribution):
|
|||
def loc(self):
|
||||
"""
|
||||
Return the location of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the loc parameter of the distribution.
|
||||
"""
|
||||
return self._loc
|
||||
|
||||
|
|
@ -108,6 +131,9 @@ class Gumbel(TransformedDistribution):
|
|||
def scale(self):
|
||||
"""
|
||||
Return the scale of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the scale parameter of the distribution.
|
||||
"""
|
||||
return self._scale
|
||||
|
||||
|
|
@ -155,7 +181,8 @@ class Gumbel(TransformedDistribution):
|
|||
.. math::
|
||||
STD(X) = \frac{\pi}{\sqrt(6)} * scale
|
||||
"""
|
||||
scale = self.scale * self.fill(self.parameter_type, self.broadcast_shape, 1.0)
|
||||
scale = self.scale * \
|
||||
self.fill(self.parameter_type, self.broadcast_shape, 1.0)
|
||||
return scale * np.pi / self.sqrt(self.const(6.))
|
||||
|
||||
def _entropy(self):
|
||||
|
|
@ -165,7 +192,8 @@ class Gumbel(TransformedDistribution):
|
|||
.. math::
|
||||
H(X) = 1. + \log(scale) + Euler-Mascheroni_constant
|
||||
"""
|
||||
scale = self.scale * self.fill(self.parameter_type, self.broadcast_shape, 1.0)
|
||||
scale = self.scale * \
|
||||
self.fill(self.parameter_type, self.broadcast_shape, 1.0)
|
||||
return 1. + self.log(scale) + np.euler_gamma
|
||||
|
||||
def _log_prob(self, value):
|
||||
|
|
@ -219,8 +247,9 @@ class Gumbel(TransformedDistribution):
|
|||
loc_b = self.cast(loc_b, self.parameter_type)
|
||||
scale_b = self.cast(scale_b, self.parameter_type)
|
||||
return self.log(scale_b / self.scale) +\
|
||||
np.euler_gamma * (self.scale / scale_b - 1.) + (self.loc - loc_b) / scale_b +\
|
||||
self.expm1((loc_b - self.loc) / scale_b + self.lgamma(self.scale / scale_b + 1.))
|
||||
np.euler_gamma * (self.scale / scale_b - 1.) + (self.loc - loc_b) / scale_b +\
|
||||
self.expm1((loc_b - self.loc) / scale_b +
|
||||
self.lgamma(self.scale / scale_b + 1.))
|
||||
|
||||
def _sample(self, shape=()):
|
||||
shape = self.checktuple(shape, 'shape')
|
||||
|
|
|
|||
|
|
@ -23,10 +23,18 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class LogNormal(msd.TransformedDistribution):
|
||||
"""
|
||||
r"""
|
||||
LogNormal distribution.
|
||||
A log-normal (or lognormal) distribution is a continuous probability distribution of a random variable whose
|
||||
logarithm is normally distributed. It is constructed as the exponential transformation of a Normal distribution.
|
||||
logarithm is normally distributed.
|
||||
The log-normal distribution has the range :math:`(0, \inf)` with the pdf as
|
||||
|
||||
.. math::
|
||||
f(x, \mu, \sigma) = 1 / x\sigma\sqrt{2\pi} \exp(-(\ln(x) - \mu)^2 / 2\sigma^2).
|
||||
|
||||
where :math:`\mu, \sigma` are the mean and
|
||||
the standard deviation of the underlying normal distribution respectively.
|
||||
It is constructed as the exponential transformation of a Normal distribution.
|
||||
|
||||
Args:
|
||||
loc (int, float, list, numpy.ndarray, Tensor): The mean of the underlying Normal distribution. Default: None.
|
||||
|
|
@ -36,6 +44,19 @@ class LogNormal(msd.TransformedDistribution):
|
|||
dtype (mindspore.dtype): type of the distribution. Default: mstype.float32.
|
||||
name (str): the name of the distribution. Default: 'LogNormal'.
|
||||
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is 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`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -44,6 +65,10 @@ class LogNormal(msd.TransformedDistribution):
|
|||
`dist_spec_args` are `loc` and `scale`.
|
||||
`dtype` must be a float type because LogNormal 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
|
||||
|
|
@ -82,7 +107,7 @@ class LogNormal(msd.TransformedDistribution):
|
|||
|
||||
self.log_2pi = np.log(2 * np.pi)
|
||||
|
||||
#ops needed for the class
|
||||
# ops needed for the class
|
||||
self.dtypeop = P.DType()
|
||||
self.exp = exp_generic
|
||||
self.expm1 = P.Expm1()
|
||||
|
|
@ -103,6 +128,9 @@ class LogNormal(msd.TransformedDistribution):
|
|||
"""
|
||||
Distribution parameter for the pre-transformed mean
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the loc parameter of the distribution.
|
||||
"""
|
||||
return self._loc
|
||||
|
||||
|
|
@ -111,6 +139,9 @@ class LogNormal(msd.TransformedDistribution):
|
|||
"""
|
||||
Distribution parameter for the pre-transformed standard deviation
|
||||
after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the scale parameter of the distribution.
|
||||
"""
|
||||
return self._scale
|
||||
|
||||
|
|
|
|||
|
|
@ -24,16 +24,35 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class Logistic(Distribution):
|
||||
"""
|
||||
r"""
|
||||
Logistic distribution.
|
||||
A Logistic distributio is a continuous distribution with the range :math:`(-\inf, \inf)`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / b \exp(\exp(-(x - a) / b) - x).
|
||||
|
||||
where a and b are loc and scale parameter respectively.
|
||||
|
||||
Args:
|
||||
loc (int, float, list, numpy.ndarray, Tensor): The location of the Logistic distribution. Default: None.
|
||||
scale (int, float, list, numpy.ndarray, Tensor): The scale of the Logistic distribution. Default: None.
|
||||
loc (float, list, numpy.ndarray, Tensor): The location of the Logistic distribution. Default: None.
|
||||
scale (float, list, numpy.ndarray, Tensor): The scale of the Logistic distribution. Default: None.
|
||||
seed (int): The seed used in sampling. The global seed is used if it is None. Default: None.
|
||||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Logistic'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is 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`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -42,6 +61,10 @@ class Logistic(Distribution):
|
|||
`dist_spec_args` are `loc` and `scale`.
|
||||
`dtype` must be a float type because Logistic distributions are continuous.
|
||||
|
||||
Raises:
|
||||
ValueError: When scale <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -127,7 +150,8 @@ class Logistic(Distribution):
|
|||
param = dict(locals())
|
||||
param['param_dict'] = {'loc': loc, 'scale': scale}
|
||||
valid_dtype = mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
super(Logistic, self).__init__(seed, dtype, name, param)
|
||||
|
||||
self._loc = self._add_parameter(loc, 'loc')
|
||||
|
|
@ -184,6 +208,9 @@ class Logistic(Distribution):
|
|||
def loc(self):
|
||||
"""
|
||||
Return the location of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the loc parameter of the distribution.
|
||||
"""
|
||||
return self._loc
|
||||
|
||||
|
|
@ -191,6 +218,9 @@ class Logistic(Distribution):
|
|||
def scale(self):
|
||||
"""
|
||||
Return the scale of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the scale parameter of the distribution.
|
||||
"""
|
||||
return self._scale
|
||||
|
||||
|
|
|
|||
|
|
@ -25,8 +25,16 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class Normal(Distribution):
|
||||
"""
|
||||
r"""
|
||||
Normal distribution.
|
||||
A Normal distributio is a continuous distribution with the range :math:`(-\inf, \inf)`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, \mu, \sigma) = 1 / \sigma\sqrt{2\pi} \exp(-(x - \mu)^2 / 2\sigma^2).
|
||||
|
||||
where :math:`\mu, \sigma` are the mean and
|
||||
the standard deviation of the normal distribution respectively.
|
||||
|
||||
Args:
|
||||
mean (int, float, list, numpy.ndarray, Tensor): The mean of the Normal distribution. Default: None.
|
||||
|
|
@ -35,6 +43,18 @@ class Normal(Distribution):
|
|||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Normal'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is 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`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -43,6 +63,10 @@ class Normal(Distribution):
|
|||
`dist_spec_args` are `mean` and `sd`.
|
||||
`dtype` must be a float type because Normal distributions are continuous.
|
||||
|
||||
Raises:
|
||||
ValueError: When sd <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
|
|||
|
|
@ -25,8 +25,11 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class Poisson(Distribution):
|
||||
"""
|
||||
r"""
|
||||
Poisson Distribution.
|
||||
A Poisson Distribution is a discrete distribution with the range as the non-negative integers,
|
||||
and the probability mass function as :math:`P(X = k) = \lambda^k \exp(-\lambda) / k!, k = 1, 2, ...`,
|
||||
where :math:`\lambda` is the rate of the distribution.
|
||||
|
||||
Args:
|
||||
rate (list, numpy.ndarray, Tensor): The rate of the Poisson distribution. Default: None.
|
||||
|
|
@ -34,6 +37,18 @@ class Poisson(Distribution):
|
|||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Poisson'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is 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`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend``
|
||||
|
||||
|
|
@ -41,6 +56,10 @@ class Poisson(Distribution):
|
|||
`rate` must be strictly greater than 0.
|
||||
`dist_spec_args` is `rate`.
|
||||
|
||||
Raises:
|
||||
ValueError: When rate <= 0.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -119,7 +138,8 @@ class Poisson(Distribution):
|
|||
param = dict(locals())
|
||||
param['param_dict'] = {'rate': rate}
|
||||
valid_dtype = mstype.int_type + mstype.uint_type + mstype.float_type
|
||||
Validator.check_type_name("dtype", dtype, valid_dtype, type(self).__name__)
|
||||
Validator.check_type_name(
|
||||
"dtype", dtype, valid_dtype, type(self).__name__)
|
||||
|
||||
# As some operators can't accept scalar input, check the type here
|
||||
if isinstance(rate, (int, float)):
|
||||
|
|
@ -151,6 +171,9 @@ class Poisson(Distribution):
|
|||
def rate(self):
|
||||
"""
|
||||
Return `rate` of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the rate parameter of the distribution.
|
||||
"""
|
||||
return self._rate
|
||||
|
||||
|
|
|
|||
|
|
@ -28,18 +28,36 @@ class TransformedDistribution(Distribution):
|
|||
Transformed Distribution.
|
||||
This class contains a bijector and a distribution and transforms the original distribution
|
||||
to a new distribution through the operation defined by the bijector.
|
||||
If X is an random variable following the underying distribution,
|
||||
and g(x) is a function represented by the bijector,
|
||||
then Y = g(X) is a random variable following the transformed distribution.
|
||||
|
||||
Args:
|
||||
bijector (Bijector): The transformation to perform.
|
||||
distribution (Distribution): The original distribution. Must has a float dtype.
|
||||
seed (int): The seed is used in sampling. The global seed is used if it is None. Default:None.
|
||||
distribution (Distribution): The original distribution. Must be a float dtype.
|
||||
seed (int): The seed is used in sampling. The global seed is used if it is None. Default: None.
|
||||
If this seed is given when a TransformedDistribution object is initialized, the object's sampling function
|
||||
will use this seed; elsewise, the underlying distribution's seed will be used.
|
||||
name (str): The name of the transformed distribution. Default: 'transformed_distribution'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`
|
||||
- `mean`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
Raises:
|
||||
TypeError: When the input `bijector` is not a Bijector instance.
|
||||
TypeError: When the input `distribution` is not a Distribution instance.
|
||||
|
||||
Note:
|
||||
The arguments used to initialize the original distribution cannot be None.
|
||||
For example, mynormal = msd.Normal(dtype=mindspore.float32) cannot be used to initialized a
|
||||
|
|
@ -93,8 +111,10 @@ class TransformedDistribution(Distribution):
|
|||
[nn.probability.bijector.Bijector], type(self).__name__)
|
||||
validator.check_value_type('distribution', distribution,
|
||||
[Distribution], type(self).__name__)
|
||||
validator.check_type_name("dtype", distribution.dtype, mstype.float_type, type(self).__name__)
|
||||
super(TransformedDistribution, self).__init__(seed, distribution.dtype, name, param)
|
||||
validator.check_type_name(
|
||||
"dtype", distribution.dtype, mstype.float_type, type(self).__name__)
|
||||
super(TransformedDistribution, self).__init__(
|
||||
seed, distribution.dtype, name, param)
|
||||
|
||||
self._bijector = bijector
|
||||
self._distribution = distribution
|
||||
|
|
@ -121,21 +141,44 @@ class TransformedDistribution(Distribution):
|
|||
# broadcast bijector batch_shape and distribution batch_shape
|
||||
self._broadcast_shape = self._broadcast_bijector_dist()
|
||||
|
||||
|
||||
@property
|
||||
def bijector(self):
|
||||
"""
|
||||
Return the bijector of the transformed distribution.
|
||||
|
||||
Output:
|
||||
Bijector, the bijector of the transformed distribution.
|
||||
"""
|
||||
return self._bijector
|
||||
|
||||
@property
|
||||
def distribution(self):
|
||||
"""
|
||||
Return the underlying distribution of the transformed distribution.
|
||||
|
||||
Output:
|
||||
Bijector, the underlying distribution of the transformed distribution.
|
||||
"""
|
||||
return self._distribution
|
||||
|
||||
@property
|
||||
def dtype(self):
|
||||
"""
|
||||
Return the dtype of the transformed distribution.
|
||||
|
||||
Output:
|
||||
Mindspore.dtype, the dtype of the transformed distribution.
|
||||
"""
|
||||
return self._dtype
|
||||
|
||||
@property
|
||||
def is_linear_transformation(self):
|
||||
"""
|
||||
Return whether the transformation is linear.
|
||||
|
||||
Output:
|
||||
Bool, true if the transformation is linear, and false otherwise.
|
||||
"""
|
||||
return self._is_linear_transformation
|
||||
|
||||
def _broadcast_bijector_dist(self):
|
||||
|
|
@ -144,7 +187,8 @@ class TransformedDistribution(Distribution):
|
|||
"""
|
||||
if self.batch_shape is None or self.bijector.batch_shape is None:
|
||||
return None
|
||||
bijector_shape_tensor = self.fill_base(self.dtype, self.bijector.batch_shape, 0.0)
|
||||
bijector_shape_tensor = self.fill_base(
|
||||
self.dtype, self.bijector.batch_shape, 0.0)
|
||||
dist_shape_tensor = self.fill_base(self.dtype, self.batch_shape, 0.0)
|
||||
return (bijector_shape_tensor + dist_shape_tensor).shape
|
||||
|
||||
|
|
@ -174,12 +218,14 @@ class TransformedDistribution(Distribution):
|
|||
\log(Py(a)) = \log(Px(g^{-1}(a))) + \log((g^{-1})'(a))
|
||||
"""
|
||||
inverse_value = self.bijector("inverse", value)
|
||||
unadjust_prob = self.distribution("log_prob", inverse_value, *args, **kwargs)
|
||||
unadjust_prob = self.distribution(
|
||||
"log_prob", inverse_value, *args, **kwargs)
|
||||
log_jacobian = self.bijector("inverse_log_jacobian", value)
|
||||
isneginf = self.equal_base(unadjust_prob, -np.inf)
|
||||
isnan = self.equal_base(unadjust_prob + log_jacobian, np.nan)
|
||||
return self.select_base(isneginf,
|
||||
self.select_base(isnan, unadjust_prob + log_jacobian, unadjust_prob),
|
||||
self.select_base(
|
||||
isnan, unadjust_prob + log_jacobian, unadjust_prob),
|
||||
unadjust_prob + log_jacobian)
|
||||
|
||||
def _prob(self, value, *args, **kwargs):
|
||||
|
|
|
|||
|
|
@ -24,8 +24,15 @@ from ._utils.custom_ops import exp_generic, log_generic
|
|||
|
||||
|
||||
class Uniform(Distribution):
|
||||
"""
|
||||
Example class: Uniform Distribution.
|
||||
r"""
|
||||
Uniform Distribution.
|
||||
A Uniform distributio is a continuous distribution with the range :math:`[a, b]`
|
||||
and the probability density function:
|
||||
|
||||
.. math::
|
||||
f(x, a, b) = 1 / b \exp(\exp(-(x - a) / b) - x),
|
||||
|
||||
where a and b are the lower and upper bound respectively.
|
||||
|
||||
Args:
|
||||
low (int, float, list, numpy.ndarray, Tensor): The lower bound of the distribution. Default: None.
|
||||
|
|
@ -34,6 +41,18 @@ class Uniform(Distribution):
|
|||
dtype (mindspore.dtype): The type of the event samples. Default: mstype.float32.
|
||||
name (str): The name of the distribution. Default: 'Uniform'.
|
||||
|
||||
Inputs and Outputs of APIs:
|
||||
The accessible api is defined in the base class, including:
|
||||
|
||||
- `prob`, `log_prob`, `cdf`, `log_cdf`, `survival_function`, and `log_survival`
|
||||
- `mean`, `sd`, `var`, and `entropy`
|
||||
- `kl_loss` and `cross_entropy`
|
||||
- `sample`
|
||||
|
||||
It should be notice that the input should be always a tensor.
|
||||
For more details of all APIs, including the inputs and outputs,
|
||||
please refer to :class:`mindspore.nn.probability.bijector.Distribution`, and examples below.
|
||||
|
||||
Supported Platforms:
|
||||
``Ascend`` ``GPU``
|
||||
|
||||
|
|
@ -42,6 +61,11 @@ class Uniform(Distribution):
|
|||
`dist_spec_args` are `high` and `low`.
|
||||
`dtype` must be float type because Uniform distributions are continuous.
|
||||
|
||||
Raises:
|
||||
ValueError: When high <= low.
|
||||
TypeError: When the input `dtype` is not a subclass of float.
|
||||
|
||||
|
||||
Examples:
|
||||
>>> import mindspore
|
||||
>>> import mindspore.nn as nn
|
||||
|
|
@ -181,13 +205,19 @@ class Uniform(Distribution):
|
|||
def low(self):
|
||||
"""
|
||||
Return the lower bound of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the lower bound of the distribution.
|
||||
"""
|
||||
return self._low
|
||||
|
||||
@property
|
||||
def high(self):
|
||||
"""
|
||||
Return the upper bound of the distribution after casting to dtype..
|
||||
Return the upper bound of the distribution after casting to dtype.
|
||||
|
||||
Output:
|
||||
Tensor, the upper bound of the distribution.
|
||||
"""
|
||||
return self._high
|
||||
|
||||
|
|
|
|||
Loading…
Reference in New Issue