Markov chain Monte Carlo samplers. More...

Classes
class	adaptive_cdhmc
	Adaptive "constant distance" Hamiltonian Monte Carlo (CDHMC) sampler. More...

class	adaptive_hmc
	Adaptive Hamiltonian Monte Carlo (HMC) sampler. More...

class	adaptive_sampler
	An abstract base class for adaptive samplers. More...

class	chains
	An `mcmc::chains` object stores parameter names and dimensionalities along with samples from multiple chains. More...

class	DualAverage
	Implements Nesterov's dual average algorithm. More...

class	hmc
	Hamiltonian Monte Carlo (HMC) sampler. More...

class	hmc_base

class	nuts
	No-U-Turn Sampler (NUTS). More...

class	nuts_diag
	No-U-Turn Sampler (NUTS) with varying step sizes. More...

class	sample
	Representation of a MCMC sample. More...

Functions
const std::vector< std::string > &	test_match_return_names (const std::vector< std::string > &names, const std::vector< std::vector< size_t > > &dimss)

void	validate_prob (double p)

void	validate_dims_idxs (const std::vector< size_t > &dims, const std::vector< size_t > &idxs)
	Validate the specified indexes with respect to the specified dimensions. More...

size_t	get_offset (const std::vector< size_t > &dims, const std::vector< size_t > &idxs)
	Return the offset in last-index major indexing for the specified indexes given the specified number of dimensions. More...

void	increment_indexes (const std::vector< size_t > &dims, std::vector< size_t > &idxs)
	Increments the specified indexes to refer to the next value in an array given by the specified dimensions. More...

template<class RNG >
void	permutation (std::vector< size_t > &x, size_t n, RNG &rng)
	Write a permutation into the specified vector of the specified size using the specified Boost random number generator. More...

template<typename T >
void	permute (const std::vector< size_t > &pi, const std::vector< T > &x_from, std::vector< T > &x_to)
	Write the specified permutation of the first vector into the second vector. More...

void	read_variables (const std::string filename, const size_t skip, std::vector< std::string > &names, std::vector< std::vector< size_t > > &dimss)
	Reads variable names and dims from a csv output file. More...

template<typename RNG >
size_t	add_chain (stan::mcmc::chains< RNG > &chains, const size_t chain, const std::string filename, const size_t skip)
	Adds a chain from a csv file. More...

void	write_error_msgs (std::ostream *error_msgs, const std::domain_error &e)

double	leapfrog (stan::model::prob_grad &model, std::vector< int > z, std::vector< double > &x, std::vector< double > &m, std::vector< double > &g, double epsilon, std::ostream error_msgs=0, std::ostream output_msgs=0)
	Computes the log probability for a single leapfrog step in Hamiltonian Monte Carlo. More...

double	rescaled_leapfrog (stan::model::prob_grad &model, std::vector< int > z, const std::vector< double > &step_sizes, std::vector< double > &x, std::vector< double > &m, std::vector< double > &g, double epsilon, std::ostream error_msgs=0, std::ostream output_msgs=0)

int	sample_unnorm_log (std::vector< double > probs, boost::uniform_01< boost::mt19937 & > &rand_uniform_01)

Detailed Description

Markov chain Monte Carlo samplers.

Function Documentation

◆ add_chain()

template<typename RNG >

size_t stan::mcmc::add_chain	(	stan::mcmc::chains< RNG > &	chains,
		const size_t	chain,
		const std::string	filename,
		const size_t	skip
	)

Adds a chain from a csv file.

Parameters

[in,out]	chains	The chains object to modify
	chain	chain number
	filename	file name of a csv output file
	skip	number of variables to skip

Returns: number of samples added

Definition at line 1788 of file chains.hpp.

◆ get_offset()

size_t stan::mcmc::get_offset	(	const std::vector< size_t > &	dims,
		const std::vector< size_t > &	idxs
	)

Return the offset in last-index major indexing for the specified indexes given the specified number of dimensions.

If both sequences are empty, the index returned is 0.

Parameters

dims	Sequence of dimensions.
idxs	Sequence of inndexes.

Returns: Offset of indexes given dimensions.

Exceptions

std::invalid_argument	If the sizes of the index and dimension sequences is different.
std::out_of_range	If one of the indexes is greater than or equal to the corresponding index.

Definition at line 103 of file chains.hpp.

◆ increment_indexes()

void stan::mcmc::increment_indexes	(	const std::vector< size_t > &	dims,
		std::vector< size_t > &	idxs
	)

Increments the specified indexes to refer to the next value in an array given by the specified dimensions.

The indexing is in last-index major order, which is column-major for matrices.

The first index in the sequence is all zeroes. Incrementing the last index, whose values are the dimensions minus one, returns the all-zero matrix.

Given dims == (2,2,2), the sequence of indexes are

[0 0 0], [1 0 0], [0 1 0], [1 1 0], [0 0 1], [1 0 1], [0 1 1], [1 1 1], [0 0 0], [1 0 0], ...

Parameters

dims	Dimensions of array.
idxs	Indexes into array.

Exceptions

std::invalid_argument	If the dimensions and indexes are not the same size.
std::out_of_range	If an index is greater than or equal to the corresponding dimension.

Definition at line 151 of file chains.hpp.

◆ leapfrog()

double stan::mcmc::leapfrog	(	stan::model::prob_grad &	model,
		std::vector< int >	z,
		std::vector< double > &	x,
		std::vector< double > &	m,
		std::vector< double > &	g,
		double	epsilon,
		std::ostream *	error_msgs = `0`,
		std::ostream *	output_msgs = `0`
	)

Computes the log probability for a single leapfrog step in Hamiltonian Monte Carlo.

If a domain error occurs when calling the model's grad_log_prob(), this function returns -inf. Domain errors can occur when distribution functions are called with parameters out of support.

Parameters

[in]	model	Probability model with gradients.
[in]	z	Integer parameters.
[in]	x	Real parameters
[in,out]	m	Momentum.
[in,out]	g	Gradient at x, z.
[in]	epsilon	Step size used in Hamiltonian dynamics.
[in,out]	error_msgs	Output stream for error messages.
[in,out]	output_msgs	Output stream for output messages.

Returns: the log probability of x and m.

Definition at line 57 of file util.hpp.

◆ permutation()

template<class RNG >

void stan::mcmc::permutation	(	std::vector< size_t > &	x,
		size_t	n,
		RNG &	rng
	)

Write a permutation into the specified vector of the specified size using the specified Boost random number generator.

The vector will be resized to the specified size.

Template Parameters

RNG	Type of random number geneation engine

Parameters

x	Vector into which to write the permutation
n	Size of permutation to create
rng	Random-number generator.

Definition at line 173 of file chains.hpp.

◆ permute()

template<typename T >

void stan::mcmc::permute	(	const std::vector< size_t > &	pi,
		const std::vector< T > &	x_from,
		std::vector< T > &	x_to
	)

Write the specified permutation of the first vector into the second vector.

The second vector will be resized to the size of the permutation.

Template Parameters

T	Type of elements to permute

Parameters

pi	Permutation.
x_from	Vector of elements to permute
x_to	Vector into which permutation of elements is written

Exceptions

std::invalid_argument If the permutation vector and source vector from which to copy are not the same size.

Definition at line 203 of file chains.hpp.

◆ read_variables()

void stan::mcmc::read_variables	(	const std::string	filename,
		const size_t	skip,
		std::vector< std::string > &	names,
		std::vector< std::vector< size_t > > &	dimss
	)

Reads variable names and dims from a csv output file.

Parameters

[in]	filename	Name of a csv output file.
[in]	skip	Number of variables to skip
[out]	names	Names of the variables
[out]	dimss	Dimensions of the variables

Definition at line 1759 of file chains.hpp.

◆ rescaled_leapfrog()

double stan::mcmc::rescaled_leapfrog	(	stan::model::prob_grad &	model,
		std::vector< int >	z,
		const std::vector< double > &	step_sizes,
		std::vector< double > &	x,
		std::vector< double > &	m,
		std::vector< double > &	g,
		double	epsilon,
		std::ostream *	error_msgs = `0`,
		std::ostream *	output_msgs = `0`
	)

Definition at line 79 of file util.hpp.

◆ sample_unnorm_log()

int stan::mcmc::sample_unnorm_log	(	std::vector< double >	probs,
		boost::uniform_01< boost::mt19937 & > &	rand_uniform_01
	)

Definition at line 103 of file util.hpp.

◆ test_match_return_names()

const std::vector<std::string>& stan::mcmc::test_match_return_names	(	const std::vector< std::string > &	names,
		const std::vector< std::vector< size_t > > &	dimss
	)

Definition at line 38 of file chains.hpp.

◆ validate_dims_idxs()

void stan::mcmc::validate_dims_idxs	(	const std::vector< size_t > &	dims,
		const std::vector< size_t > &	idxs
	)

Validate the specified indexes with respect to the specified dimensions.

Parameters

dims	Dimensions of array.
idxs	Indexes into array.

Exceptions

std::invalid_argument	If the two arrays are different sizes.
std::out_of_range	If any of the indexes is greater than or equal to its correpsonding dimension.

Definition at line 70 of file chains.hpp.

◆ validate_prob()

void stan::mcmc::validate_prob ( double p )

Definition at line 49 of file chains.hpp.

◆ write_error_msgs()

void stan::mcmc::write_error_msgs	(	std::ostream *	error_msgs,
		const std::domain_error &	e
	)

Definition at line 19 of file util.hpp.

Classes

Functions

Detailed Description

Function Documentation

◆ add_chain()

◆ get_offset()

◆ increment_indexes()

◆ leapfrog()

◆ permutation()

◆ permute()

◆ read_variables()

◆ rescaled_leapfrog()

◆ sample_unnorm_log()

◆ test_match_return_names()

◆ validate_dims_idxs()

◆ validate_prob()

◆ write_error_msgs()