skgpuppy.UncertaintyPropagation module

class skgpuppy.UncertaintyPropagation.UncertaintyPropagation

Bases: object

propagate(y, u, Sigma_x)

Propagates the uncertain density Girard2004 (page 32)

Parameters:
  • y – point to estimate the output density at
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

output density
propagate_many(yvec, u, Sigma_x)

Propagates the uncertain density Girard2004 (page 32)

Parameters:
  • yvec – vector of points to estimate the output density at
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

output densities
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationApprox(gp)

Bases: skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA

Parameters:gp – callable gaussian process that returns mean and variance for a given input vector x
propagate_GA(u, Sigma_x)
propagate_mean(u, Sigma_x)
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationExact(gp)

Bases: skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA

Parameters:gp – callable gaussian process that returns mean and variance for a given input vector x
propagate_GA(u, Sigma_x)
propagate_mean(u, Sigma_x, C_ux=None)
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA(gp)

Bases: object

Superclass for all UncertaintyPropagationGA Classes

Parameters:gp – callable gaussian process that returns mean and variance for a given input vector x
propagate_GA(u, Sigma_x)

Propagates the uncertainty using the gaussian approximation from Girard2004

Parameters:
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

mean, variance of the output
propagate_mean(u, Sigma_x)

Propagates the mean using the gaussian approximation from Girard2004

Parameters:
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

mean of the output
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationLinear(gp)

Bases: skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA

Parameters:gp – callable gaussian process that returns mean and variance for a given input vector x
propagate_GA(u, Sigma_x)
propagate_mean(u, Sigma_x)
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationMC(gp, n=1000)

Bases: skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA, skgpuppy.UncertaintyPropagation.UncertaintyPropagation

Using Monte Carlo Integration -> Very inefficient but very stable

Parameters:
  • gp – callable gaussian process that returns mean and variance for a given input vector x
  • n – number of samples
propagate(y, u, Sigma_x)
propagate_GA(u, Sigma_x)
propagate_many(yvec, u, Sigma_x)

Propagates the uncertain density Girard2004 (page 32)

Parameters:
  • yvec – vector of points to estimate the output density at
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

output densities
propagate_mean(u, Sigma_x)
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationNumerical(gp)

Bases: skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA, skgpuppy.UncertaintyPropagation.UncertaintyPropagation

The numerical propagation works fine if we want predictions for the noisy f But it is unstable for small variances

Deprecated since version Use: UncertaintyPropagationNumericalHG instead

Parameters:gp – callable gaussian process that returns mean and variance for a given input vector x
propagate(y, u, Sigma_x)
propagate_GA(u, Sigma_x)
propagate_many(yvec, u, Sigma_x)

Propagates the uncertain density Girard2004 (page 32)

Parameters:
  • yvec – vector of points to estimate the output density at
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

output densities
propagate_mean(u, Sigma_x)
class skgpuppy.UncertaintyPropagation.UncertaintyPropagationNumericalHG(gp)

Bases: skgpuppy.UncertaintyPropagation.UncertaintyPropagationGA, skgpuppy.UncertaintyPropagation.UncertaintyPropagation

The numerical propagation works fine if we want predictions for the noisy f

Parameters:gp – callable gaussian process that returns mean and variance for a given input vector x
propagate(y, u, Sigma_x)
propagate_GA(u, Sigma_x)
propagate_many(yvec, u, Sigma_x)

Propagates the uncertain density Girard2004 (page 32)

Parameters:
  • yvec – vector of points to estimate the output density at
  • u – vector of means
  • Sigma_x – covariance Matrix of the input
Returns:

output densities
propagate_mean(u, Sigma_x)