Tools for defininig and running multilevel and multiindex MCMC algorithms.

Multiindex MCMC methods are built on an entire arbitriry-dimensional grid of sampling problems, in contrast to classical MCMC only sampling from a single distribution.

In order to be effective, MIMCMC methods require distributions closely related to each other, where evaluating the coarser ones should be significantly computationally cheaper. A typical example are models based on the Finite Element method, where strongly varying mesh resolutions lead to a hierarchy of models. Then, coarser models are a good approximation of finer ones while being far cheaper to compute.

In analogy to the mathematical definition, running a MIMCMC method requires defining a grid of models (as well as other needed components) via MIComponentFactory. Most importantly, how proposals will be drawn from coarser chains has to be defined as well as how to combine them (via MIInterpolation) with a proposal for the current fine chain.

Refer to the MCMC examples for complete code examples.

Classes
class	muq::SamplingAlgorithms::MIMCMC
	Multiindex MCMC method. More...

class	muq::SamplingAlgorithms::DefaultComponentFactory
	Provides a high level interface for the sampling problems on each MIMCMC level. More...

class	muq::SamplingAlgorithms::MIComponentFactory
	Interface defining models on a multiindex structure. More...

Classes