TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions
¶
Module Contents¶
Classes¶
Posterior distribution of the conductivity field of a Poisson problem. |
|
Posterior distribution of the conductivity field of a Poisson problem. |
- class TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions.Covariance[source]¶
Bases:
object
- class TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions.OrnsteinUhlenbeckCovariance(var, l)[source]¶
Bases:
Covariance
- class TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions.SquaredExponentialCovariance(var, l)[source]¶
Bases:
Covariance
- class TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions.IdentityCovariance(var)[source]¶
Bases:
Covariance
- class TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions.GaussianFieldPoissonDistribution(solver, sens_pos_list, sens_geo_std, prior_mean_field_vals=None, prior_cov=OrnsteinUhlenbeckCovariance(1.0, 1.0), lkl_cov=IdentityCovariance(1.0), field_vals=None, real_observations=None)[source]¶
Bases:
TransportMaps.Distributions.Inference.BayesPosteriorDistribution
Posterior distribution of the conductivity field of a Poisson problem.
The system is observed through the operator \(\mathcal{O}[{\bf c}]\) defined by
\[\mathcal{O}[{\bf c}](u) = \int u({\bf x}) \frac{1}{2\pi\sigma_y^2} \exp\left(-\frac{\Vert{\bf x} - {\bf c}\Vert^2}{2\sigma_y^2} \right) d{\bf x} \;,\]where \(u\) is the solution associated to an underlying field \(\kappa\).
Given the sensor locations \(\{{\bf c}_i\}_{i=1}^s\), the Bayesian problem is defined as follows:
\[\begin{split}{\bf y}_i = \mathcal{O}[{\bf c}_i](u) + \varepsilon \;, \qquad \varepsilon \sim \mathcal{GP}(0, \mathcal{C}_y({\bf x},{\bf x}^\prime)) \\ \kappa({\bf x}) \sim \log\mathcal{GP}(\mu_\kappa({\bf x}), \mathcal{C}_\kappa({\bf x},{\bf x}^\prime))\end{split}\]- Parameters:
solver (PoissonSolver) – Poisson solver for a particular problem setting
sens_geo_std (float) – observation kernel width \(\sigma_y\)
real_observations (
ndarray
) – observations \(\{{\bf y}_i\}_{i=1}^s\)field_vals (
ndarray
) – generating field (default isNone
, generating a synthetic field)prior_mean_field_vals (
ndarray
) – values \(\mu_\kappa({\bf x})\) (default isNone
, which corresponds to \(\mu_\kappa({\bf x})=0\))prior_cov (Covariance) – covariance function \(\mathcal{C}_\kappa({\bf x},{\bf x}^\prime)\)
lkl_cov (Covariance) – covariance function \(\mathcal{C}_y({\bf x},{\bf x}^\prime)\)
- class TransportMaps.Distributions.Examples.PoissonProblem.PoissonDistributions.GaussianFieldIndependentLikelihoodPoissonDistribution(solver, sens_pos_list, sens_geo_std, prior_mean_field_vals=None, prior_cov=OrnsteinUhlenbeckCovariance(1.0, 1.0), lkl_std=1.0, field_vals=None, real_observations=None)[source]¶
Bases:
TransportMaps.Distributions.Inference.BayesPosteriorDistribution
Posterior distribution of the conductivity field of a Poisson problem.
The system is observed through the operator \(\mathcal{O}[{\bf c}]\) defined by
\[\mathcal{O}[{\bf c}](u) = \int u({\bf x}) \frac{1}{2\pi\sigma_y^2} \exp\left(-\frac{\Vert{\bf x} - {\bf c}\Vert^2}{2\sigma_y^2} \right) d{\bf x} \;,\]where \(u\) is the solution associated to an underlying field \(\kappa\).
Given the sensor locations \(\{{\bf c}_i\}_{i=1}^s\), the Bayesian problem is defined as follows:
\[\begin{split}{\bf y}_i = \mathcal{O}[{\bf c}_i](u) + \varepsilon \;, \qquad \varepsilon \sim \mathcal{N}(0, \sigma^2 {\bf I}) \\ \kappa({\bf x}) \sim \log\mathcal{GP}(\mu_\kappa({\bf x}), \mathcal{C}_\kappa({\bf x},{\bf x}^\prime))\end{split}\]- Parameters:
solver (PoissonSolver) – Poisson solver for a particular problem setting
sens_geo_std (float) – observation kernel width \(\sigma_y\)
real_observations (
ndarray
) – observations \(\{{\bf y}_i\}_{i=1}^s\)field_vals (
ndarray
) – generating field (default isNone
, generating a synthetic field)prior_mean_field_vals (
ndarray
) – values \(\mu_\kappa({\bf x})\) (default isNone
, which corresponds to \(\mu_\kappa({\bf x})=0\))prior_cov (Covariance) – covariance function \(\mathcal{C}_\kappa({\bf x},{\bf x}^\prime)\)
lkl_std (floag) – standard deviation \(\sigma\)