skgpuppy.FFNI module

class skgpuppy.FFNI.FullFactorialNumericalIntegrationEvans(func, mean)

Bases: skgpuppy.FFNI.PropagateMoments

Class to perform error propagation using Evans Method (1967).

Warning

This method is very inaccurate (especially for skewness and kurtosis)

Parameters:
  • func – (n-d) function to approximate
  • mean – propagate uncertainty around this mean vector
propagate(Sigma_x, skew=False, kurtosis=False)

Propagates a normal distributed uncertainty around self.mean through the deterministic function self.func.

Parameters:
  • Sigma_x – Covariance matrix (assumed to be diagonal)
  • skew – Return the skewness of the resulting distribution
  • kurtosis – Return the kurtosis of the resulting distribution
Returns:

mean, variance, [skewness, [kurtosis]]
class skgpuppy.FFNI.FullFactorialNumericalIntegrationHermGauss(func, mean, order)

Bases: skgpuppy.FFNI.PropagateMoments

Class to perform error propagation using Gauss-Hermite quadrature

Parameters:
  • func – (n-d) function to approximate
  • mean – propagate uncertainty around this mean vector
  • order – order of the integration series
propagate(Sigma_x, skew=False, kurtosis=False)

Propagates a normal distributed uncertainty around self.mean through the deterministic function self.func.

Parameters:
  • Sigma_x – Covariance matrix (assumed to be diagonal)
  • skew – Return the skewness of the resulting distribution
  • kurtosis – Return the kurtosis of the resulting distribution
Returns:

mean, variance, [skewness, [kurtosis]]
class skgpuppy.FFNI.FullFactorialNumericalIntegrationNaive(func, mean)

Bases: skgpuppy.FFNI.PropagateMoments

Class to perform error propagation using Scipy numerical integration

Parameters:
  • func – (n-d) function to approximate
  • mean – propagate uncertainty around this mean vector
propagate(Sigma_x, skew=False, kurtosis=False)

Propagates a normal distributed uncertainty around self.mean through the deterministic function self.func.

Parameters:
  • Sigma_x – Covariance matrix (assumed to be diagonal)
  • skew – Return the skewness of the resulting distribution
  • kurtosis – Return the kurtosis of the resulting distribution
Returns:

mean, variance, [skewness, [kurtosis]]
class skgpuppy.FFNI.PropagateMoments(func, mean)

Bases: object

Superclass for uncertainty propagation through deterministic functions

Parameters:
  • func – (n-d) function to approximate
  • mean – propagate uncertainty around this mean vector
propagate(Sigma_x, skew=False, kurtosis=False)

Propagates a normal distributed uncertainty around self.mean through the deterministic function self.func.

Parameters:
  • Sigma_x – Covariance matrix (assumed to be diagonal)
  • skew – Return the skewness of the resulting distribution
  • kurtosis – Return the kurtosis of the resulting distribution
Returns:

mean, variance, [skewness, [kurtosis]]