Implementation of the Metropolis-Hastings algorithm. More...

#include <metropolis.hpp>

Inheritance diagram for vmcmc::MetropolisHastings:

Collaboration diagram for vmcmc::MetropolisHastings:

Classes
struct	ChainConfig
	Private class encapsulating Sample chains for parallel tempering. More...

Public Member Functions
virtual void	Initialize () override

virtual void	Advance (size_t nSteps=1) override

virtual void	Finalize () override

virtual size_t	NumberOfChains () override

virtual const Chain &	GetChain (size_t cIndex=0) override

void	SetNumberOfChains (size_t nChains)

template<typename ContainerT = std::initializer_list<double>>
void	SetBetas (ContainerT betas)

const std::vector< double > &	GetBetas () const

template<typename ProposalT , typename... ArgsT>
void	SetProposalFunction (ArgsT &&...args)

void	SetProposalFunction (std::shared_ptr< Proposal > proposalFunction)

std::shared_ptr< Proposal >	GetProposalFunction ()

std::shared_ptr< const Proposal >	GetProposalFunction () const

void	SetRandomizeStartPoint (bool randomizeStartPoint)

bool	IsRandomizeStartPoint () const

void	SetMultiThreading (bool enable)

bool	IsMultiThreading () const

double	GetSwapAcceptanceRate (size_t iChain, ptrdiff_t iBeta=-1) const
	Get the fraction of accepted swaps between tempered chains. More...

Public Member Functions inherited from vmcmc::Algorithm
void	SetParameterConfig (const ParameterConfig &paramConfig)
	Set the parameter configuration. More...

const ParameterConfig &	GetParameterConfig () const

void	SetPrior (DefaultCallable prior)

template<size_t NParams, typename CallableT >
void	SetPrior (CallableT prior)

void	SetLikelihood (DefaultCallable likelihood)

template<size_t NParams, typename CallableT >
void	SetLikelihood (CallableT likelihood)

void	SetNegLogLikelihood (DefaultCallable negLoglikelihood)

template<size_t NParams, typename CallableT >
void	SetNegLogLikelihood (CallableT negLoglikelihood)

void	SetTotalLength (size_t length)

size_t	GetTotalLength () const

template<typename WriterT , typename... ArgsT>
void	AddWriter (ArgsT &&...args)
	Add an output writer by specifying its type and passing constructor arguments. More...

void	AddWriter (std::shared_ptr< Writer > writer)

double	EvaluatePrior (const std::vector< double > &paramValues) const
	Evaluate the prior for the given parameter values. More...

double	EvaluateLikelihood (const std::vector< double > &paramValues) const
	Evaluate the target function likelihood for the given parameter values. More...

double	EvaluateNegLogLikelihood (const std::vector< double > &paramValues) const
	Evaluate the target function -log(likelihood) for the given parameter values. More...

bool	Evaluate (Sample &sample) const
	Evalutate the target function prior, likelihood and -log(likelihood) at the position defined by the `sample`, and update the `sample` accordingly. More...

void	Run ()
	Start sampling!

ChainSetStatistics &	GetStatistics ()

const ChainSetStatistics &	GetStatistics () const

Static Public Member Functions
static double	CalculateMHRatio (const Sample &prevState, const Sample &nextState, double proposalAsymmetry=1.0, double beta=1.0)

Protected Member Functions
void	AdvanceChainConfig (size_t iChainConfig, size_t iBeta, size_t nSteps=1)

void	ProposePtSwapping (size_t iChainConfig)

Protected Attributes
bool	fRandomizeStartPoint

std::vector< double >	fBetas

std::shared_ptr< Proposal >	fProposalFunction

size_t	fPtFrequency

std::vector< std::unique_ptr< ChainConfig > >	fChainConfigs

Protected Attributes inherited from vmcmc::Algorithm
ParameterConfig	fParameterConfig

DefaultCallable	fPrior

DefaultCallable	fLikelihood

DefaultCallable	fNegLogLikelihood

size_t	fTotalLength

size_t	fCycleLength

std::vector< std::shared_ptr< Writer > >	fWriters

ChainSetStatistics	fStatistics

Additional Inherited Members
Public Types inherited from vmcmc::Algorithm
using	DefaultCallable = std::function< double(const std::vector< double > &)>

Detailed Description

Implementation of the Metropolis-Hastings algorithm.

The transition kernel alias proposal function can be defined by the user according to the interface of #Proposal. By default, a multivariate Gaussian is used, based on the parameter configuration.

Parallel tempering (useful for multimodal likelihoods) can be activated by specifying additional values for beta < 1 (reciprocal temperature), thus constructing flatter distributions. For each value of beta, an individual sampling chain, together with it's own parameter configuration and proposal function is set up.

Member Function Documentation

void vmcmc::MetropolisHastings::Advance ( size_t nSteps = 1 )

overridevirtual

In case of multi-core parallelization, all active chains (number of chain sets * number of PT beta values) are progressed in parallel by nSteps each.

Implements vmcmc::Algorithm.

double vmcmc::MetropolisHastings::GetSwapAcceptanceRate	(	size_t	iChain,
		ptrdiff_t	iBeta = `-1`
	)		const

Get the fraction of accepted swaps between tempered chains.

Parameters

iChain	Index of the chain set.
iBeta	Index of the tempered chain pair. If < 0, average over all pairs.

Returns

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

vmcmc/metropolis.hpp
vmcmc/metropolis.cpp

Classes

Public Member Functions

Static Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Member Function Documentation