RandomVariable¶
-
class
probnum.random_variables.
RandomVariable
(shape: Union[numbers.Integral, Iterable[numbers.Integral]], dtype: Union[numpy.dtype, str], random_state: Union[None, int, numpy.random.mtrand.RandomState, numpy.random._generator.Generator] = None, parameters: Optional[Dict[str, Any]] = None, sample: Optional[Callable[Tuple[int, ...], ValueType]] = None, in_support: Optional[Callable[ValueType, bool]] = None, cdf: Optional[Callable[ValueType, numpy.float64]] = None, logcdf: Optional[Callable[ValueType, numpy.float64]] = None, quantile: Optional[Callable[numbers.Real, ValueType]] = None, mode: Optional[Callable[ValueType]] = None, median: Optional[Callable[ValueType]] = None, mean: Optional[Callable[ValueType]] = None, cov: Optional[Callable[ValueType]] = None, var: Optional[Callable[ValueType]] = None, std: Optional[Callable[ValueType]] = None, entropy: Optional[Callable[numpy.float64]] = None, as_value_type: Optional[Callable[Any, ValueType]] = None)¶ Bases:
typing.Generic
Random variables are the main objects used by probabilistic numerical methods.
Every probabilistic numerical method takes a random variable encoding the prior distribution as input and outputs a random variable whose distribution encodes the uncertainty arising from finite computation. The generic signature of a probabilistic numerical method is:
output_rv = probnum_method(input_rv, method_params)
In practice, most random variables used by methods in ProbNum have Dirac or Gaussian measure.
Instances of
RandomVariable
can be added, multiplied, etc. with arrays and linear operators. This may change theirdistribution
and not necessarily all previously available methods are retained.The internals of
RandomVariable
objects are assumed to be constant over their whole lifecycle. This is due to the caches used to make certain computations more efficient. As a consequence, altering the internal state of aRandomVariable
(e.g. its mean, cov, sampling function, etc.) will result in undefined behavior. In particular, this should be kept in mind when subclassingRandomVariable
or any of its descendants.Parameters: - shape – Shape of realizations of this random variable.
- dtype – Data type of realizations of this random variable. If
object
will be converted tonumpy.dtype
. - as_value_type –
Function which can be used to transform user-supplied arguments, interpreted as realizations of this random variable, to an easy-to-process, normalized format. Will be called internally to transform the argument of functions like
in_support
,cdf
andlogcdf
,pmf
andlogpmf
(inDiscreteRandomVariable
),pdf
andlogpdf
(inContinuousRandomVariable
), and potentially by similar functions in subclasses.For instance, this method is useful if (
log
)``cdf`` and (log
)``pdf`` both only work onnp.float_
arguments, but we still want the user to be able to pass Pythonfloat
. Thenas_value_type
should be set to something likelambda x: np.float64(x)
.
See also
asrandvar
- Transform into a
RandomVariable
.
Examples
Attributes Summary
T
Transpose the random variable. cov
Covariance \(\operatorname{Cov}(X) = \mathbb{E}((X-\mathbb{E}(X))(X-\mathbb{E}(X))^\top)\) of the random variable. dtype
Data type of (elements of) a realization of this random variable. entropy
mean
Mean \(\mathbb{E}(X)\) of the distribution. median
Median of the random variable. median_dtype
The dtype of the median
.mode
Mode of the random variable. moment_dtype
The dtype of any (function of a) moment of the random variable, e.g. ndim
parameters
Parameters of the probability distribution. random_state
Random state of the random variable. shape
Shape of realizations of the random variable. size
std
Standard deviation of the distribution. var
Variance \(\operatorname{Var}(X) = \mathbb{E}((X-\mathbb{E}(X))^2)\) of the distribution. Methods Summary
cdf
(x)Cumulative distribution function. in_support
(x)infer_median_dtype
(value_dtype, str])infer_moment_dtype
(value_dtype, str])logcdf
(x)Log-cumulative distribution function. quantile
(p)Quantile function. reshape
(newshape, Iterable[numbers.Integral]])Give a new shape to a random variable. sample
(size, Iterable[numbers.Integral]] = ())Draw realizations from a random variable. transpose
(*axes)Transpose the random variable. Attributes Documentation
-
T
¶ Transpose the random variable.
Parameters: axes (None, tuple of ints, or n ints) – See documentation of numpy.ndarray.transpose. Returns: transposed_rv Return type: The transposed random variable.
-
cov
¶ Covariance \(\operatorname{Cov}(X) = \mathbb{E}((X-\mathbb{E}(X))(X-\mathbb{E}(X))^\top)\) of the random variable.
To learn about the dtype of the covariance, see
moment_dtype
.Returns: cov – The kernels of the random variable. Return type: array-like
-
dtype
¶ Data type of (elements of) a realization of this random variable.
-
entropy
¶
-
mean
¶ Mean \(\mathbb{E}(X)\) of the distribution.
To learn about the dtype of the mean, see
moment_dtype
.Returns: mean – The mean of the distribution. Return type: array-like
-
median
¶ Median of the random variable.
To learn about the dtype of the median, see
median_dtype
.Returns: median – The median of the distribution. Return type: float
-
median_dtype
¶ The dtype of the
median
. It will be set to the dtype arising from the multiplication of values with dtypesdtype
andnp.float_
. This is motivated by the fact that, even for discrete random variables, e.g. integer-valued random variables, themedian
might lie in between two values in which case these values are averaged. For example, a uniform random variable on \(\{ 1, 2, 3, 4 \}\) will have a median of \(2.5\).
-
mode
¶ Mode of the random variable.
Returns: mode – The mode of the random variable. Return type: float
-
moment_dtype
¶ The dtype of any (function of a) moment of the random variable, e.g. its
mean
,cov
,var
, orstd
. It will be set to the dtype arising from the multiplication of values with dtypesdtype
andnp.float_
. This is motivated by the mathematical definition of a moment as a sum or an integral over products of probabilities and values of the random variable, which are represented as using the dtypesnp.float_
anddtype
, respectively.
-
ndim
¶
-
parameters
¶ Parameters of the probability distribution.
The parameters of the distribution such as mean, variance, et cetera stored in a
dict
.
-
random_state
¶ Random state of the random variable.
This attribute defines the RandomState object to use for drawing realizations from this random variable. If None (or np.random), the global np.random state is used. If integer, it is used to seed the local
RandomState
instance.
-
shape
¶ Shape of realizations of the random variable.
-
size
¶
-
std
¶ Standard deviation of the distribution.
To learn about the dtype of the standard deviation, see
moment_dtype
.Returns: std – The standard deviation of the distribution. Return type: array-like
-
var
¶ Variance \(\operatorname{Var}(X) = \mathbb{E}((X-\mathbb{E}(X))^2)\) of the distribution.
To learn about the dtype of the variance, see
moment_dtype
.Returns: var – The variance of the distribution. Return type: array-like
Methods Documentation
-
cdf
(x: ValueType) → numpy.float64[source]¶ Cumulative distribution function.
Parameters: x (array-like) – Evaluation points of the cumulative distribution function. The shape of this argument should be (..., S1, ..., SN)
, where(S1, ..., SN)
is theshape
of the random variable. The cdf evaluation will be broadcast over all additional dimensions.Returns: q – Value of the cumulative density function at the given points. Return type: array-like
-
logcdf
(x: ValueType) → numpy.float64[source]¶ Log-cumulative distribution function.
Parameters: x (array-like) – Evaluation points of the cumulative distribution function. The shape of this argument should be (..., S1, ..., SN)
, where(S1, ..., SN)
is theshape
of the random variable. The logcdf evaluation will be broadcast over all additional dimensions.Returns: q – Value of the log-cumulative density function at the given points. Return type: array-like
-
quantile
(p: numbers.Real) → ValueType[source]¶ Quantile function.
The quantile function \(Q \colon [0, 1] \to \mathbb{R}\) of a random variable \(X\) is defined as \(Q(p) = \inf\{ x \in \mathbb{R} \colon p \le F_X(x) \}\), where \(F_X \colon \mathbb{R} \to [0, 1]\) is the
cdf()
of the random variable. From the definition it follows that the quantile function always returns values of the same dtype as the random variable. For instance, for a discrete distribution over the integers, the returned quantiles will also be integers. This means that, in general, \(Q(0.5)\) is not equal to themedian
as it is defined in this class. See https://en.wikipedia.org/wiki/Quantile_function for more details and examples.
-
reshape
(newshape: Union[numbers.Integral, Iterable[numbers.Integral]]) → probnum.random_variables.RandomVariable[source]¶ Give a new shape to a random variable.
Parameters: newshape (int or tuple of ints) – New shape for the random variable. It must be compatible with the original shape. Returns: reshaped_rv Return type: self
with the new dimensions ofshape
.
-
sample
(size: Union[numbers.Integral, Iterable[numbers.Integral]] = ()) → ValueType[source]¶ Draw realizations from a random variable.
Parameters: size (tuple) – Size of the drawn sample of realizations. Returns: sample – Sample of realizations with the given size
and the inherentshape
.Return type: array-like