TaylorExpansionMoments¶

class TaylorExpansionMoments(*args)¶

First and second order Taylor expansion formulas.

Refer to Taylor decomposition importance factors.

Parameters

limitStateVariableRandomVector: It must be of type Composite, which means it must have been defined with the class CompositeRandomVector.

Notes

In a probabilistic approach the Taylor expansion can be used propagate the uncertainties of the input variables $\uX$ through the model $h$ towards the output variables $\uY$ . It enables to access the central dispersion (Expectation, Variance) of the output variables.

This method is based on a Taylor decomposition of the output variable $\uY$ towards the $\uX$ random vectors around the mean point $\muX$ . Depending on the order of the Taylor decomposition (classically first order or second order), one can obtain different formulas introduced hereafter.

As $\uY=h(\uX)$ , the Taylor decomposition around $\ux = \muX$ at the second order yields to:

$\uY = h(\muX) + <\vect{\vect{\nabla}}h(\muX) , \: \uX - \muX> + \frac{1}{2}<<\vect{\vect{\vect{\nabla }}}^2 h(\muX,\: \vect{\mu}_{\:X}),\: \uX - \muX>,\: \uX - \muX> + o(\Cov \uX)$

where:

$\muX = \Expect{\uX}$ is the vector of the input variables at the mean values of each component.
$\Cov \uX$ is the covariance matrix of the random vector uX. The elements are the followings : $(\Cov \uX)_{ij} = \Expect{\left(X^i - \Expect{X^i} \right)^2}$
$\vect{\vect{\nabla}} h(\muX) = \: \Tr{\left( \frac{\partial y^i}{\partial x^j}\right)}_{\ux\: =\: \muX} = \: \Tr{\left( \frac{\partial h^i(\ux)}{\partial x^j}\right)}_{\ux\: =\: \muX}$ is the transposed Jacobian matrix with $i=1,\ldots,n_Y$ and $j=1,\ldots,n_X$ .
$\vect{\vect{\vect{\nabla^2}}} h(\ux\:,\ux)$ is a tensor of order 3. It is composed by the second order derivative towards the $i^\textrm{th}$ and $j^\textrm{th}$ components of $\ux$ of the $k^\textrm{th}$ component of the output vector $h(\ux)$ . It yields to: $\left( \nabla^2 h(\ux) \right)_{ijk} = \frac{\partial^2 (h^k(\ux))}{\partial x^i \partial x^j}$
$<\vect{\vect{\nabla}}h(\muX) , \: \uX - \muX> = \sum_{j=1}^{n_X} \left( \frac{\partial {\uy}}{\partial {x^j}}\right)_{\ux = \muX} . \left( X^j-\muX^j \right)$
$<<\vect{\vect{\vect{\nabla }}}^2 h(\muX,\: \vect{\mu}_{X}),\: \uX - \muX>,\: \uX - \muX> = \left( \Tr{(\uX^i - \muX^i)}. \left(\frac{\partial^2 y^k}{\partial x^i \partial x^k}\right)_{\ux = \muX}. (\uX^j - \muX^j) \right)_{ijk}$

Approximation at the order 1:

Expectation:

$\Expect{\uY} \approx \vect{h}(\muX)$

Pay attention that $\Expect{\uY}$ is a vector. The $k^\textrm{th}$ component of this vector is equal to the $k^\textrm{th}$ component of the output vector computed by the model $h$ at the mean value. $\Expect{\uY}$ is thus the computation of the model at mean.

Variance:

$\Cov \uY \approx \Tr{\vect{\vect{\nabla}}}\:\vect{h}(\muX).\Cov \uX.\vect{\vect{\nabla}}\:\vect{h}(\muX)$

Approximation at the order 2:

Expectation:

$(\Expect{\uY})_k \approx (\vect{h}(\muX))_k + \left( \sum_{i=1}^{n_X}\frac{1}{2} (\Cov \uX)_{ii}.{(\nabla^2\:h(\uX))}_{iik} + \sum_{i=1}^{n_X} \sum_{j=1}^{i-1} (\Cov X)_{ij}.{(\nabla^2\:h(\uX))}_{ijk} \right)_k$

Variance:

The decomposition of the variance at the order 2 is not implemented. It requires both the knowledge of higher order derivatives of the model and the knowledge of moments of order strictly greater than 2 of the PDF.

Examples

>>> import openturns as ot
>>> ot.RandomGenerator.SetSeed(0)
>>> myFunc = ot.SymbolicFunction(['x1', 'x2', 'x3', 'x4'],
...     ['(x1*x1+x2^3*x1)/(2*x3*x3+x4^4+1)', 'cos(x2*x2+x4)/(x1*x1+1+x3^4)'])
>>> R = ot.CorrelationMatrix(4)
>>> for i in range(4):
...     R[i, i - 1] = 0.25
>>> distribution = ot.Normal([0.2]*4, [0.1, 0.2, 0.3, 0.4], R)
>>> # We create a distribution-based RandomVector
>>> X = ot.RandomVector(distribution)
>>> # We create a composite RandomVector Y from X and myFunc
>>> Y = ot.CompositeRandomVector(myFunc, X)
>>> # We create a Taylor expansion method to approximate moments
>>> myTaylorExpansionMoments = ot.TaylorExpansionMoments(Y)
>>> print(myTaylorExpansionMoments.getMeanFirstOrder())
[0.0384615,0.932544]

Methods

`drawImportanceFactors`()	Draw the importance factors.
`getClassName`()	Accessor to the object's name.
`getCovariance`()	Get the approximation at the first order of the covariance matrix.
`getGradientAtMean`()	Get the gradient of the function.
`getHessianAtMean`()	Get the hessian of the function.
`getId`()	Accessor to the object's id.
`getImportanceFactors`()	Get the importance factors.
`getLimitStateVariable`()	Get the limit state variable.
`getMeanFirstOrder`()	Get the approximation at the first order of the mean.
`getMeanSecondOrder`()	Get the approximation at the second order of the mean.
`getName`()	Accessor to the object's name.
`getShadowedId`()	Accessor to the object's shadowed id.
`getValueAtMean`()	Get the value of the function.
`getVisibility`()	Accessor to the object's visibility state.
`hasName`()	Test if the object is named.
`hasVisibleName`()	Test if the object has a distinguishable name.
`setName`(name)	Accessor to the object's name.
`setShadowedId`(id)	Accessor to the object's shadowed id.
`setVisibility`(visible)	Accessor to the object's visibility state.

__init__(*args)¶

drawImportanceFactors()¶

Draw the importance factors.

Returns

graphGraph: Graph containing the pie corresponding to the importance factors of the probabilistic variables.

getClassName()¶

Accessor to the object’s name.

Returns

class_namestr: The object class name (object.__class__.__name__).

getCovariance()¶

Get the approximation at the first order of the covariance matrix.

Returns

covarianceCovarianceMatrix: Approximation at the first order of the covariance matrix of the random vector.

getGradientAtMean()¶

Get the gradient of the function.

Returns

gradientMatrix: Gradient of the Function which defines the random vector at the mean point of the input random vector.

getHessianAtMean()¶

Get the hessian of the function.

Returns

hessianSymmetricTensor: Hessian of the Function which defines the random vector at the mean point of the input random vector.

getId()¶

Accessor to the object’s id.

Returns

idint: Internal unique identifier.

getImportanceFactors()¶

Get the importance factors.

Returns

factorsPoint: Importance factors of the inputs : only when randVect is of dimension 1.

getLimitStateVariable()¶

Get the limit state variable.

Returns

limitStateVariableRandomVector: Limit state variable.

getMeanFirstOrder()¶

Get the approximation at the first order of the mean.

Returns

meanPoint: Approximation at the first order of the mean of the random vector.

getMeanSecondOrder()¶

Get the approximation at the second order of the mean.

Returns

meanPoint: Approximation at the second order of the mean of the random vector (it requires that the hessian of the Function has been defined).

getName()¶

Accessor to the object’s name.

Returns

namestr: The name of the object.

getShadowedId()¶

Accessor to the object’s shadowed id.

Returns

idint: Internal unique identifier.

getValueAtMean()¶

Get the value of the function.

Returns

valuePoint: Value of the Function which defines the random vector at the mean point of the input random vector.

getVisibility()¶

Accessor to the object’s visibility state.

Returns

visiblebool: Visibility flag.

hasName()¶

Test if the object is named.

Returns

hasNamebool: True if the name is not empty.

hasVisibleName()¶

Test if the object has a distinguishable name.

Returns

hasVisibleNamebool: True if the name is not empty and not the default one.

setName(name)¶

Accessor to the object’s name.

Parameters

namestr: The name of the object.

setShadowedId(id)¶

Accessor to the object’s shadowed id.

Parameters

idint: Internal unique identifier.

setVisibility(visible)¶

Accessor to the object’s visibility state.

Parameters

visiblebool: Visibility flag.

OpenTURNS

An Open source initiative for the Treatment of Uncertainties, Risks'N Statistics

Previous topic

Next topic

This Page

TaylorExpansionMoments¶