MorrisSensitivity

class MorrisSensitivity(randomParameters=False)

Class to define the Morris sensitivity benchmark problem.

Methods

getFirstOrderIndices()

Returns the first order Sobol' sensitivity indices.

getFunction()

Returns the function.

getInputDistribution()

Returns the input distribution.

getName()

Returns the name of the problem.

getTotalOrderIndices()

Returns the total order Sobol' sensitivity indices.

__init__(randomParameters=False)

Define the Morris sensitivity benchmark problem.

The function g is defined on [0,1]^{20} and takes values in \mathbb{R}.

Its inputs are independent random variables with distribution \mathcal{U}(0,1).

It is defined as:

g(\boldsymbol{x}) & = \beta_0
       + \sum_{i} \beta_i\, w_i(x) \\
     & \quad + \sum_{i<j} \beta_{i,j}\, w_i(x)\, w_j(x) \\
     & \quad + \sum_{i<j<k} \beta_{i,j,k}\, w_i(x)\, w_j(x)\, w_k(x) \\
     & \quad + \sum_{i<j<k<\ell} \beta_{i,j,k,\ell}\, w_i(x)\, w_j(x)\, w_k(x)\, w_\ell(x)

for any \boldsymbol{x} \in [0,1]^{20} where:

w_i(x) = 2\,(x_i - 0.5)

for i \in \{1, 2, 4, 6, 8, \dots, 20\}, and:

w_i(x) = 2\,\left(\frac{1.1\, x_i}{x_i + 1} - 0.5\right)

for i \in \{3, 5, 7\}. The parameters (\beta_i)_{1 \leq i \leq 20}, (\beta_{i,j})_{1 \leq i, j \leq 20}, (\beta_{i,j, k})_{1 \leq i, j, k\leq 20}, (\beta_{i,j, k, \ell})_{1 \leq i, j, k, \ell \leq 20} are presented below, in the Notes section. In (Morris, 1991)’s paper, these parameters are random. In order to get consistent results, the default value of the randomParameters parameter is so that the parameters beta are constant, deterministic, values.

Parameters:
randomParameters: bool

Set to True to get random parameters. Caution: the reference Sobol’ indices are not valid if the parameters are random.

Returns:
None.

Notes

The dimension of this problem is equal to 20 and cannot be changed.

The function is the sum of five functions.

  • The first function is constant and equal to \beta_0.

  • The second is a linear combination of \boldsymbol{w} coefficients.

  • The third is an order 2 combination of \boldsymbol{w} coefficients.

  • The fourth is an order 3 combination of \boldsymbol{w} coefficients.

  • The fifth is an order 4 combination of \boldsymbol{w} coefficients.

Therefore, Morris’s function is a modification of an order 4 polynomial, where the small non-linearity comes from the \boldsymbol{w} coefficients.

The reference Sobol’ indices were computed from polynomial chaos expansion. A Sobol’ low discrepancy design of experiments was generated with 65536 training points. The full polynomial chaos expansion used an hyperbolic enumeration rule (using quasi-norm parameter 0.7) and a polynomial degree 6. The coefficients were estimated from least squares. With 65536 points in the validation set, the Q² was 99.47%. There are 3 significant digits in the reference results.

Morris Function Parameters

Let \mathcal{N}(0,1) be a random Gaussian variable with zero mean and unit standard deviation.

The first order coefficients of the Morris function are:

\beta_i =
\begin{cases}
20 & \text{if } i = 1,\ldots,10, \\
\mathcal{N}(0,1) & \text{otherwise}.
\end{cases}

The second order coefficients are:

\beta_{i,j} =
\begin{cases}
-15 & \text{if } i,j \in \{1,\ldots,6\}, \\
\mathcal{N}(0,1) & \text{otherwise}.
\end{cases}

The third order coefficients are:

\beta_{i,j,\ell} =
\begin{cases}
-10 & \text{if } i,j,\ell \in \{1,\ldots,5\},\\
0 & \text{otherwise}.
\end{cases}

The fourth order coefficients are:

\beta_{i,j,\ell,s} =
\begin{cases}
5 & \text{if } i,j,\ell,s \in \{1,\ldots,4\},\\
0 & \text{otherwise}.
\end{cases}

References

  • M. D. Morris, 1991, Factorial sampling plans for preliminary computational experiments, Technometrics, 33, 161–174.

getFirstOrderIndices()

Returns the first order Sobol’ sensitivity indices.

Parameters:
None.
Returns:
firstOrderIndices: ot.Point

The first order sensitivity indices.

getFunction()

Returns the function.

Parameters:
None.
Returns:
function: ot.Function

The function.

getInputDistribution()

Returns the input distribution.

Parameters:
None.
Returns:
distribution: ot.Distribution

The distribution.

getName()

Returns the name of the problem.

Parameters:
None.
Returns:
name: str

The name.

getTotalOrderIndices()

Returns the total order Sobol’ sensitivity indices.

Parameters:
None.
Returns:
totalOrderIndices: ot.Point

The total order sensitivity indices.