An implemental of the Adaptive Metropolis algorithm. More...

#include <AMProposal.h>

Inheritance diagram for muq::SamplingAlgorithms::AMProposal:

Detailed Description

An implemental of the Adaptive Metropolis algorithm.

Configuration Parameters:

Parameter Key	Type	Default Value	Description
"ProposalVariance"	Double	1.0	Initial proposal variance before adaptation begins.
"AdaptSteps"	Integer	-	How often the proposal covariance should be updated.
"AdaptStart"	Integer	-	How many steps should be taken before beginning to adapt the covariance.
"AdaptScale"	Double	-	Scaling of the sample covariance used to define the proposal covariance: \(\Sigma_{prop} = \alpha \hat{\Sigma}\)

Definition at line 21 of file AMProposal.h.

Public Member Functions
	AMProposal (boost::property_tree::ptree pt, std::shared_ptr< AbstractSamplingProblem > const &prob)

	AMProposal (boost::property_tree::ptree pt, std::shared_ptr< AbstractSamplingProblem > const &prob, Eigen::MatrixXd const &initialCov)

virtual	~AMProposal ()=default

virtual void	Adapt (unsigned int const t, std::vector< std::shared_ptr< SamplingState >> const &states) override
	Adapt the proposal after each step. More...

virtual Eigen::MatrixXd	ProposalCovariance () const

Public Member Functions inherited from muq::SamplingAlgorithms::MCMCProposal
	MCMCProposal (boost::property_tree::ptree const &pt, std::shared_ptr< AbstractSamplingProblem > const &probIn)

virtual	~MCMCProposal ()=default

void	SetCommunicator (std::shared_ptr< parcer::Communicator > newcomm)

virtual void	SetBlockInd (int newBlockInd)

virtual int	GetBlockInd () const

Additional Inherited Members
Public Types inherited from muq::SamplingAlgorithms::MCMCProposal
typedef std::function< std::shared_ptr< MCMCProposal >boost::property_tree::ptree, std::shared_ptr< AbstractSamplingProblem >)>	MCMCProposalConstructor

typedef std::map< std::string, MCMCProposalConstructor >	MCMCProposalMap

Static Public Member Functions inherited from muq::SamplingAlgorithms::MCMCProposal
static std::shared_ptr< MCMCProposal >	Construct (boost::property_tree::ptree const &pt, std::shared_ptr< AbstractSamplingProblem > const &probIn)
	Static constructor for the transition kernel. More...

static std::shared_ptr< MCMCProposalMap >	GetMCMCProposalMap ()

Constructor & Destructor Documentation

◆ AMProposal() [1/2]

muq::SamplingAlgorithms::AMProposal::AMProposal	(	boost::property_tree::ptree	pt,
		std::shared_ptr< AbstractSamplingProblem > const &	prob
	)

◆ AMProposal() [2/2]

muq::SamplingAlgorithms::AMProposal::AMProposal	(	boost::property_tree::ptree	pt,
		std::shared_ptr< AbstractSamplingProblem > const &	prob,
		Eigen::MatrixXd const &	initialCov
	)

Construct the adaptive Metropolis proposal with a specific estimate of the target covariance matrix. Note that initailCov will be scaled by the adaptScale parameter set in the ptree and should therefore be an estiamte of the target covariance, not a specification of the proposal covariance.

◆ ~AMProposal()

virtual muq::SamplingAlgorithms::AMProposal::~AMProposal ( )

virtualdefault

Member Function Documentation

◆ Adapt()

void AMProposal::Adapt	(	unsigned int const	t,
		std::vector< std::shared_ptr< SamplingState >> const &	states
	)

overridevirtual

Adapt the proposal after each step.

Adapt the proposal covariance.

Parameters

[in]	t	The current step
[in]	state	The current state

Reimplemented from muq::SamplingAlgorithms::MCMCProposal.

Reimplemented in muq::SamplingAlgorithms::ParallelAMProposal.

Definition at line 43 of file AMProposal.cpp.

References adaptEnd, adaptStart, adaptSteps, muq::SamplingAlgorithms::MCMCProposal::blockInd, mean, newSamps, numAdaptSamps, numNewSamps, and propChol.

Referenced by muq::SamplingAlgorithms::ParallelAMProposal::Adapt().

◆ ConstructCovariance()

Eigen::MatrixXd AMProposal::ConstructCovariance	(	boost::property_tree::ptree const &	pt,
		std::shared_ptr< AbstractSamplingProblem > const &	prob
	)

staticprivate

Definition at line 34 of file AMProposal.cpp.

References muq::SamplingAlgorithms::MCMCProposal::prob.

◆ LogDensity()

double AMProposal::LogDensity	(	std::shared_ptr< SamplingState > const &	currState,
		std::shared_ptr< SamplingState > const &	propState
	)

overrideprivatevirtual

Implements muq::SamplingAlgorithms::MCMCProposal.

Definition at line 97 of file AMProposal.cpp.

References adaptScale, muq::SamplingAlgorithms::MCMCProposal::blockInd, numAdaptSamps, and propChol.

◆ ProposalCovariance()

Eigen::MatrixXd AMProposal::ProposalCovariance ( ) const

virtual

Returns the current proposal covariance.

Definition at line 79 of file AMProposal.cpp.

References adaptScale, numAdaptSamps, and propChol.

◆ Sample()

std::shared_ptr< SamplingState > AMProposal::Sample ( std::shared_ptr< SamplingState > const & currentState )

overrideprivatevirtual

Implements muq::SamplingAlgorithms::MCMCProposal.

Definition at line 84 of file AMProposal.cpp.

References adaptScale, muq::SamplingAlgorithms::MCMCProposal::blockInd, numAdaptSamps, and propChol.

Member Data Documentation

◆ adaptEnd

const unsigned int muq::SamplingAlgorithms::AMProposal::adaptEnd

private

After how many iterations should we stop adapting?

Definition at line 85 of file AMProposal.h.

Referenced by Adapt().

◆ adaptScale

const double muq::SamplingAlgorithms::AMProposal::adaptScale

private

Definition at line 88 of file AMProposal.h.

Referenced by LogDensity(), ProposalCovariance(), and Sample().

◆ adaptStart

const unsigned int muq::SamplingAlgorithms::AMProposal::adaptStart

private

When should we start adapting?

Definition at line 82 of file AMProposal.h.

Referenced by Adapt().

◆ adaptSteps

const unsigned int muq::SamplingAlgorithms::AMProposal::adaptSteps

private

How frequently should we update the adaption?

Definition at line 79 of file AMProposal.h.

Referenced by Adapt().

◆ mean

Eigen::VectorXd muq::SamplingAlgorithms::AMProposal::mean

private

The current mean.

Definition at line 70 of file AMProposal.h.

Referenced by Adapt().

◆ newSamps

Eigen::MatrixXd muq::SamplingAlgorithms::AMProposal::newSamps

private

Samples seen between visits.

Definition at line 63 of file AMProposal.h.

Referenced by Adapt().

◆ numAdaptSamps

unsigned int muq::SamplingAlgorithms::AMProposal::numAdaptSamps =0

private

Total number of samples that have been used so far to compute mean and covariance.

Definition at line 67 of file AMProposal.h.

Referenced by Adapt(), LogDensity(), ProposalCovariance(), and Sample().

◆ numNewSamps

unsigned int muq::SamplingAlgorithms::AMProposal::numNewSamps =0

private

Definition at line 64 of file AMProposal.h.

Referenced by Adapt().

◆ propChol

Eigen::LLT<Eigen::MatrixXd, Eigen::Lower> muq::SamplingAlgorithms::AMProposal::propChol

private

The Cholesky factorization of the current proposal covariance.

Definition at line 76 of file AMProposal.h.

Referenced by Adapt(), LogDensity(), ProposalCovariance(), and Sample().

◆ propCov

Eigen::MatrixXd muq::SamplingAlgorithms::AMProposal::propCov

private

The current proposal covariance.

Definition at line 73 of file AMProposal.h.

The documentation for this class was generated from the following files:

doxygen_prep/MUQ/SamplingAlgorithms/AMProposal.h
doxygen_prep/modules/SamplingAlgorithms/src/AMProposal.cpp

Detailed Description

Public Member Functions

Additional Inherited Members

Constructor & Destructor Documentation

◆ AMProposal() [1/2]

◆ AMProposal() [2/2]

◆ ~AMProposal()

Member Function Documentation

◆ Adapt()

◆ ConstructCovariance()

◆ LogDensity()

◆ ProposalCovariance()

◆ Sample()

Member Data Documentation

◆ adaptEnd

◆ adaptScale

◆ adaptStart

◆ adaptSteps

◆ mean

◆ newSamps

◆ numAdaptSamps

◆ numNewSamps

◆ propChol

◆ propCov