Contains the current state and the history of the DREAM chains. More...

#include <tsgDreamState.hpp>

Public Member Functions
	TasmanianDREAM ()
	Constructor for a null DREAM state, no chains and no dimensions (used for MPI purposes).

	TasmanianDREAM (int cnum_chains, int cnum_dimensions)
	Constructor for a DREAM state with the number of chains and dimensions.

	TasmanianDREAM (int cnum_chains, const TasGrid::TasmanianSparseGrid &grid)
	Constructor for a DREAM state with the number of chains and the dimension of the sparse grid.

	~TasmanianDREAM ()
	Default destructor, release all used memory.

int	getNumDimensions () const
	Return the number of dimensions.

int	getNumChains () const
	Return the number of chains.

size_t	getNumHistory () const
	Return the number of saved vectors in the history.

bool	isStateReady () const
	Return true if the state has already been initialized with setState().

bool	isPDFReady () const
	Return true if the pdf values have already been initialized with setPDF().

void	setState (const std::vector< double > &new_state)
	Set the current DREAM state to the new_state. More...

void	setState (std::function< void(double *)> update_state)
	Set the current DREAM state by calling update_state on each vector. More...

void	setPDFvalues (const std::vector< double > &new_values)
	Set the current set of pdf values to the new_values. More...

void	setPDFvalues (std::function< void(const std::vector< double > &state, std::vector< double > &values)> probability_distribution)
	Set the current set of pdf values using the given probability_distribution. More...

void	clearPDFvalues ()
	Erase the currently stored pdf values, useful when switching between log and non-log form sampling.

void	getIJKdelta (size_t i, size_t j, size_t k, double w, std::vector< double > &x) const
	Used by the DREAM sampler, , where s indicates the current i, j, and k-th state vectors. More...

void	getChainState (size_t i, double *x) const
	Return the state of the i-th chain, store an array x of size at least num_dimensions. More...

const std::vector< double > &	getChainState () const
	Return a const reference to the internal state vector.

double	getPDFvalue (size_t i) const
	Return the value of the probability_distribution of the i-th chain. More...

void	expandHistory (int num_snapshots)
	Allocate (expand) internal storage for the history snapshots, avoids reallocating data when saving a snapshot. More...

void	saveStateHistory (size_t num_accepted)
	Appends the current state to the history. More...

const std::vector< double > &	getHistory () const
	Return a const reference to the internal state vector.

const std::vector< double > &	getHistoryPDF () const
	Return a const reference to the internal state vector.

void	getHistoryMeanVariance (std::vector< double > &mean, std::vector< double > &var) const
	Compute the means and variance of the saved history.

void	getApproximateMode (std::vector< double > &mode) const
	Return the sample with highest probability, searchers within the history.

std::vector< double >	getApproximateMode () const
	Overload that returns the vector.

void	clearHistory ()
	Clear the stored history (does not touch the state).

double	getAcceptanceRate () const
	Returns the acceptance rate of the current history.

Static Public Member Functions
static const char *	getVersion ()
	Return human readable string with the version.

static const char *	getLicense ()
	Return human readable sting with the license, refer to the LICENSE file for details.

static int	getVersionMajor ()
	Return the major version of the library.

static int	getVersionMinor ()
	Return the minor version of the library.

Detailed Description

Contains the current state and the history of the DREAM chains.

DREAM State: The DREAM state consists of num_chains vectors of size num_dimensions, where the dimension is either specified or inferred from the given sparse grid. Each vector is also associated with a value of the corresponding probability density function (pdf). The history of the state consists of multiple snapshots of the vectors and the pdf values.

Container class: The TasmanianDREAM class is a container for the current state and history, where the vectors are stored using std::vector data structures. Methods are provided for file I/O, direct access to the data and several handy queries for inference or optimization problems.

Member Function Documentation

◆ setState() [1/2]

void TasDREAM::TasmanianDREAM::setState ( const std::vector< double > & new_state )

Set the current DREAM state to the new_state.

The new_state must have size equal to num_chains times num_dimensions, the first chain should be stored in the first num_dimensions entries, the second chain in the second batch of num_dimensions entries, and so on.

◆ setState() [2/2]

void TasDREAM::TasmanianDREAM::setState ( std::function< void(double *)> update_state )

Set the current DREAM state by calling update_state on each vector.

The update_state function will be given each state vector as an array and the function should overwrite the entries with the desired current state. For example, update_state can overwrite the array with random numbers sampled according to a desired probability distribution.

◆ setPDFvalues() [1/2]

void TasDREAM::TasmanianDREAM::setPDFvalues ( const std::vector< double > & new_values )

Set the current set of pdf values to the new_values.

The new_values must have size equal to the number of chains, the current set of pdf values are overwritten with the new ones. Note: that there is no restriction on the sign of the values, since log-form sampling can result in negative values.

◆ setPDFvalues() [2/2]

void TasDREAM::TasmanianDREAM::setPDFvalues ( std::function< void(const std::vector< double > &state, std::vector< double > &values)> probability_distribution )

Set the current set of pdf values using the given probability_distribution.

Calls the probability_distribution function wit the current state (the state must be already initialized). The probability_distribution is the same function as in the DREAM sampling call, in order to avoid discrepancy, it may be best to leave the PDF values uninitialized and let the DREAM sampler to call setPDFvalues() automatically.

◆ getIJKdelta()

void TasDREAM::TasmanianDREAM::getIJKdelta	(	size_t	i,
		size_t	j,
		size_t	k,
		double	w,
		std::vector< double > &	x
	)		const

Used by the DREAM sampler, , where s indicates the current i, j, and k-th state vectors.

Used by the DREAM sampler and probably should not be called by the user. This returns the vector of the i-th chain corrected by the difference between the k-th and j-th chain and weighted by w. The result is written to the array x, which must have size at least num_dimensions.

◆ getChainState()

void TasDREAM::TasmanianDREAM::getChainState	(	size_t	i,
		double *	x
	)		const

inline

Return the state of the i-th chain, store an array x of size at least num_dimensions.

Used by the DREAM sampler and probably should not be called by the user.

◆ getPDFvalue()

double TasDREAM::TasmanianDREAM::getPDFvalue ( size_t i ) const

inline

Return the value of the probability_distribution of the i-th chain.

Used by the DREAM sampler and probably should not be called by the user.

◆ expandHistory()

void TasDREAM::TasmanianDREAM::expandHistory ( int num_snapshots )

Allocate (expand) internal storage for the history snapshots, avoids reallocating data when saving a snapshot.

Used by the DREAM sampler and probably should not be called by the user.

◆ saveStateHistory()

void TasDREAM::TasmanianDREAM::saveStateHistory ( size_t num_accepted )

Appends the current state to the history.

Used by the DREAM sampler and probably should not be called by the user; num_accepted is the number of new chains in the state.

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

DREAM/tsgDreamState.hpp

Public Member Functions

Static Public Member Functions

Detailed Description

Member Function Documentation

◆ setState() [1/2]

◆ setState() [2/2]

◆ setPDFvalues() [1/2]

◆ setPDFvalues() [2/2]

◆ getIJKdelta()

◆ getChainState()

◆ getPDFvalue()

◆ expandHistory()

◆ saveStateHistory()