Distributions#

Distribution classes and associated methods.

The classes in this module describe statistical distributions. Most of them are non-normalized having implications for quantifying the evidence term of Bayes’ rule.

All of the classes inherit from _AbstractDistribution; a base class outlining required methods and their signatures (required in- and outputs).

Note

A tutorial on implementing your own distributions can be found at Creating your own inverse problem.

Available classes within Distributions:

_AbstractDistribution

Abstract base class for distributions.

StandardNormal1D

Standard normal distribution in 1 dimension.

Normal

Normal distribution in model space.

Laplace

Laplace distribution in model space.

Uniform

Uniform bounded or unbouded distribution in model space.

LinearMatrix

Likelihood model based on a linear forward model given as \(G \mathbf{m} = \mathbf{d}\) coupled with Gaussian observational errors.

CompositeDistribution

Distribution distribution combined from multiple unconditional distributions.

AdditiveDistribution

Distribution generated by summing the characteristic functions of two other distributions.

BayesRule

A class to apply (the unnormalized) Bayes' rule to two or more distributions.

Himmelblau

Himmelblau's 2-dimensional function.

SourceLocation2D

Earthquake source location in 2D using a single velocity for the subsurface.

SourceLocation3D

Earthquake source location in 3D using a single velocity for the subsurface.