PythonRandomVector¶

(Source code, png)

class PythonRandomVector(dim=0)¶

Allow one to overload RandomVector from Python.

Parameters:

dimpositive int: Vector dimension. Default is 0.

See also

RandomVector

Examples

>>> import openturns as ot
>>> ot.RandomGenerator.SetSeed(0)

Overload RandomVector from Python:

>>> class RVEC(ot.PythonRandomVector):
...    def __init__(self):
...        super(RVEC, self).__init__(2)
...        self.setDescription(['R', 'S'])
...
...    def getRealization(self):
...        X = [ot.RandomGenerator.Generate(), 2 + ot.RandomGenerator.Generate()]
...        return X
...
...    def getSample(self, size):
...        X = []
...        for i in range(size):
...            X.append([ot.RandomGenerator.Generate(), 2 + ot.RandomGenerator.Generate()])
...        return X
...
...    def getMean(self):
...        return [0.5, 2.5]
...
...    def getCovariance(self):
...        return [[0.0833333, 0.], [0., 0.0833333]]

Use the overloaded class:

>>> R = RVEC()
>>> # Instance creation
>>> myRV = ot.RandomVector(R)
>>> # Realization
>>> print(myRV.getRealization())
[0.629877,2.88281]
>>> # Sample
>>> print(myRV.getSample(5))
0 : [ 0.135276  2.0325    ]
1 : [ 0.347057  2.96942   ]
2 : [ 0.92068   2.50304   ]
3 : [ 0.0632061 2.29276   ]
4 : [ 0.714382  2.38336   ]
>>> # Mean
>>> print(myRV.getMean())
[0.5,2.5]
>>> # Covariance
>>> print(myRV.getCovariance())
[[ 0.0833333 0         ]
 [ 0         0.0833333 ]]

Random vectors can admit parameters.

In the following example, we define a RandomVector to sample from a normal multivariate normal distribution truncated to a ball. We implement the setParameter method to define the ball’s center.

>>> class NormalTruncatedToBall(ot.PythonRandomVector):
...    def __init__(self, dim, max_dist):
...        super().__init__(dim)
...        self._center = ot.Point(dim)
...        self._normal = ot.Normal(dim)
...        self._max_dist = max_dist
...        self.setParameter(ot.Point(dim))
...
...    def getRealization(self):
...        dist = ot.SpecFunc.MaxScalar
...        while dist>self._max_dist:
...            candidate = self._normal.getRealization()
...            dist = (candidate - self._center).norm()
...        return candidate
...
...    def setParameter(self, center): # the parameter influences sampling
...        self._center = center
...
...    def getParameter(self): # implemented for the sake of consistency
...        return self._center
...
...    def getParameterDescription(self): # optional
...        return ["center_{}".format(i) for i in range(self.getDimension())]

Define an instance of this RandomVector and set the parameter:

>>> myRV = ot.RandomVector(NormalTruncatedToBall(2, 1.5))
>>> myRV.setParameter([1.3, 0.6])

Get a sample and plot it:

>>> sample = myRV.getSample(100)
>>> graph = ot.Graph("Sample from a PythonRandomVector", "", "", True, '')
>>> cloud = ot.Cloud(sample)
>>> graph.add(cloud)
>>> from openturns.viewer import View
>>> view = View(graph)