namespace for new Dakota utilities module More...

Classes
class	CholeskySolver
	The CholeskySolver class is used to solve linear systems with a symmetric matrix with a pivoted Cholesky decomposition. More...

class	DataScaler
	The DataScaler class computes the scaling coefficients and scales a 2D data matrix with dimensions num_samples by num_features. More...

class	LinearSolverBase
	The LinearSolverBase class serves as an API for derived solvers. More...

class	LUSolver
	The LUSolver class is used to solve linear systems with the LU decomposition. More...

class	NormalizationScaler
	Normalizes the data using max and min feature values. More...

class	NoScaler
	Leaves the data unscaled. More...

class	QRSolver
	The QRSolver class solves the linear least squares problem with a QR decomposition. More...

class	StandardizationScaler
	Standardizes the data so the each feature has zero mean and unit variance. More...

class	SVDSolver
	The SVDSolver class is used to solve linear systems with the singular value decomposition. More...

Typedefs
using	BimapMetrictypeStr = boost::bimap< METRIC_TYPE, std::string >
	alias for Boost Bimap metric type <--> string

using	SCALER_TYPE = DataScaler::SCALER_TYPE
	alias for DataScaler's SCALER_TYPE

using	BimapScalertypeStr = boost::bimap< SCALER_TYPE, std::string >
	alias for Boost Bimap scaler type <--> string

using	SOLVER_TYPE = LinearSolverBase::SOLVER_TYPE
	alias for LinearSolverBase's SOLVER_TYPE

using	BimapSolvertypeStr = boost::bimap< SOLVER_TYPE, std::string >
	alias for Boost Bimap solver type <--> string

Enumerations
enum	METRIC_TYPE { SUM_SQUARED, MEAN_SQUARED, ROOT_MEAN_SQUARED, SUM_ABS, MEAN_ABS, MAX_ABS, ABS_PERCENTAGE_ERROR, MEAN_ABS_PERCENTAGE_ERROR, R_SQUARED }
	Enumeration for supported metric types.

Functions
void	error (const std::string &msg)
	Throws a std::runtime_error based on the message argument. More...

bool	matrix_equals (const MatrixXi &A, const MatrixXi &B)
	Tests whether two Eigen MatrixXi objects are equal. More...

bool	matrix_equals (const MatrixXd &A, const MatrixXd &B, double tol)
	Tests whether two Eigen MatrixXd objects are equal, within a given tolerance. More...

bool	matrix_equals (const RealMatrix &A, const RealMatrix &B, double tol)
	Tests whether two Teuchos RealMatrix objects are equal, within a given tolerance. More...

bool	relative_allclose (const MatrixXd &A, const MatrixXd &B, const double tol)
	Tests whether two Eigen MatrixXd objects relatively equal (element-wise) within a given tolerance. More...

double	variance (const VectorXd &vec)
	Calculates the variance based on an Eigen VectorXd of double values. More...

void	populateVectorsFromFile (const std::string &filename, std::vector< VectorXd > &R, int num_datasets, int num_samples)
	Populate a collection of vectors read in a from a text file assuming data layout is one dataset per row. More...

void	populateMatricesFromFile (const std::string &filename, std::vector< MatrixXd > &S, int num_datasets, int num_vars, int num_samples)
	Populate a collection of matrices read in a from a text file assuming data layout is a "column-major" stack of num_samples by num_vars matrices. More...

int	n_choose_k (int n, int k)
	Calculate Binomial coefficient n choose k. More...

void	random_permutation (const int num_pts, const unsigned int seed, VectorXi &permutations)
	Random permutation of int array.

void	create_cv_folds (const int num_folds, const int num_pts, std::vector< VectorXi > &fold_indices, const int seed=22)
	Generate indices for cross validation folds.

MatrixXd	create_uniform_random_double_matrix (const int rows, const int cols)
	Generate a real-valued matrix of uniformly distributed random values.

MatrixXd	create_uniform_random_double_matrix (const int rows, const int cols, const unsigned int seed)
	Generate a real-valued matrix of uniformly distributed random values.

MatrixXd	create_uniform_random_double_matrix (const int rows, const int cols, const unsigned int seed, bool transform, const double low, const double high)
	Generate a real-valued matrix of uniformly distributed random values.

template<typename T >
int	num_nonzeros (const T &mat)
	Caclulate and return number of nonzero entries in vector or matrix. More...

template<typename T1 , typename T2 >
void	nonzero (const T1 &v, T2 &result)
	Create a vector of indices based on nonzero entries in input vector. More...

template<typename T1 , typename T2 >
void	append_columns (const T1 &new_cols, T2 &target)
	Append columns of input matrix to existing matrix. More...

template<typename T >
double	p_norm (const T &v, double p)
	Caclulate and return p-norm of a vector. More...

METRIC_TYPE	metric_type (const std::string &metric_name)
	Convert the metric from string to enum. More...

double	compute_metric (const VectorXd &p, const VectorXd &d, const std::string &metric_name)
	Computes the difference between prediction and data vectors. More...

std::shared_ptr< DataScaler >	scaler_factory (DataScaler::SCALER_TYPE scaler_type, const MatrixXd &unscaled_matrix)
	Free function to construct DataScaler. More...

std::shared_ptr< LinearSolverBase >	solver_factory (LinearSolverBase::SOLVER_TYPE type)
	Free function to construct LinearSolverBase. More...

Variables
static BimapMetrictypeStr	type_name_bimap
	Bimap between metric types and names. More...

static BimapScalertypeStr	type_name_bimap
	Bimap between scaler types and names. More...

static BimapSolvertypeStr	type_name_bimap
	Bimap between solver types and names. More...

Detailed Description

namespace for new Dakota utilities module

Function Documentation

◆ error()

void error ( const std::string & msg )

Throws a std::runtime_error based on the message argument.

Parameters

[in] msg The error message to throw

Referenced by compute_metric(), create_cv_folds(), and matrix_equals().

◆ matrix_equals() [1/3]

bool matrix_equals	(	const MatrixXi &	A,
		const MatrixXi &	B
	)

Tests whether two Eigen MatrixXi objects are equal.

Parameters

[in]	A	The first matrix to test
[in]	B	The second matrix to test

Returns: Whether the matrices are equal

◆ matrix_equals() [2/3]

bool matrix_equals	(	const MatrixXd &	A,
		const MatrixXd &	B,
		double	tol
	)

Tests whether two Eigen MatrixXd objects are equal, within a given tolerance.

Parameters

[in]	A	The first matrix to test
[in]	B	The second matrix to test
[in]	tol	The tolerance to use when comparing double values

Returns: Whether the matrices are equal to within tolerance

References error().

◆ matrix_equals() [3/3]

bool matrix_equals	(	const RealMatrix &	A,
		const RealMatrix &	B,
		double	tol
	)

Tests whether two Teuchos RealMatrix objects are equal, within a given tolerance.

Parameters

[in]	A	The first matrix to test
[in]	B	The second matrix to test
[in]	tol	The tolerance to use when comparing double values

Returns: Whether the matrices are equal to within tolerance

References error().

◆ relative_allclose()

bool relative_allclose	(	const MatrixXd &	A,
		const MatrixXd &	B,
		const double	tol
	)

Tests whether two Eigen MatrixXd objects relatively equal (element-wise) within a given tolerance.

Parameters

[in]	A	The first matrix to test
[in]	B	The second matrix to test
[in]	tol	The relative tolerance to use when comparing double values

Returns: Whether the matrices are relatively equal (within the tolerance)

◆ variance()

double variance ( const VectorXd & vec )

Calculates the variance based on an Eigen VectorXd of double values.

Parameters

[in] vec The vector

Returns: The calculated variance

◆ populateVectorsFromFile()

void populateVectorsFromFile	(	const std::string &	filename,
		std::vector< VectorXd > &	R,
		int	num_datasets,
		int	num_samples
	)

Populate a collection of vectors read in a from a text file assuming data layout is one dataset per row.

Parameters

[in]	filename	The file that contains the data
[out]	R	The collection of vectors to be populated
[in]	num_datasets	The number of datasets to read in
[in]	num_samples	The number of data points (e.g. function values, build points) per dataset

◆ populateMatricesFromFile()

void populateMatricesFromFile	(	const std::string &	filename,
		std::vector< MatrixXd > &	S,
		int	num_datasets,
		int	num_vars,
		int	num_samples
	)

Populate a collection of matrices read in a from a text file assuming data layout is a "column-major" stack of num_samples by num_vars matrices.

Parameters

[in]	filename	The file that contains the data
[out]	S	The collection of vectors to be populated
[in]	num_datasets	The number of datasets to read in
[in]	num_samples	The number of data points (e.g. function values, build points) per dataset (row dim)
[in]	num_vars	The number of variables per dataset (column dim)

◆ n_choose_k()

int n_choose_k	(	int	n,
		int	k
	)

Calculate Binomial coefficient n choose k.

Parameters

[in]	n	Number of elements in set
[in]	k	Number of elements in subset k where n >= k >= 0

Returns: Number of ways to choose an (unordered) subset of k elements from a fixed set of n elements

Referenced by dakota::surrogates::size_level_index_vector().

◆ num_nonzeros()

int dakota::util::num_nonzeros ( const T & mat )

Caclulate and return number of nonzero entries in vector or matrix.

Parameters

[in] mat Incoming vector or matrix

Returns: Number of nonzeros

Referenced by dakota::surrogates::compute_hyperbolic_level_indices(), dakota::surrogates::compute_hyperbolic_subdim_level_indices(), and nonzero().

◆ nonzero()

void dakota::util::nonzero	(	const T1 &	v,
		T2 &	result
	)

Create a vector of indices based on nonzero entries in input vector.

Parameters

[in]	v	Incoming vector
[out]	result	Vector having values at nonzero locations of incoming vector and value equal to ordinal of occurrence

References num_nonzeros().

Referenced by dakota::surrogates::compute_hyperbolic_level_indices().

◆ append_columns()

void dakota::util::append_columns	(	const T1 &	new_cols,
		T2 &	target
	)

Append columns of input matrix to existing matrix.

Parameters

[in]	new_cols	Incoming matrix of column vectors to append
[out]	target	Matrix to augment with appended columns

Referenced by dakota::surrogates::compute_hyperbolic_indices(), dakota::surrogates::compute_hyperbolic_level_indices(), and dakota::surrogates::compute_reduced_indices().

◆ p_norm()

double dakota::util::p_norm	(	const T &	v,
		double	p
	)

Caclulate and return p-norm of a vector.

Parameters

[in]	v	Incoming vector
[in]	p	Order or norm to compute

Returns: p-norm of incoming vector

Referenced by dakota::surrogates::compute_hyperbolic_subdim_level_indices().

◆ metric_type()

METRIC_TYPE metric_type ( const std::string & metric_name )

Convert the metric from string to enum.

Parameters

[in] metric_name metric

Returns: converted metric

References type_name_bimap.

Referenced by compute_metric().

◆ compute_metric()

double compute_metric	(	const VectorXd &	p,
		const VectorXd &	d,
		const std::string &	metric_name
	)

Computes the difference between prediction and data vectors.

Parameters

[in]	p	prediction vector.
[in]	d	data vector.
[in]	metric_name	metric to compute.

Returns: the value of the computed metric.

References error(), and metric_type().

Referenced by Dakota::compute_regression_coeffs(), and Surrogate::evaluate_metrics().

◆ scaler_factory()

std::shared_ptr< DataScaler > scaler_factory	(	DataScaler::SCALER_TYPE	scaler_type,
		const MatrixXd &	unscaled_matrix
	)

Free function to construct DataScaler.

Parameters

[in]	scaler_type	Which scaler to construct
[in]	unscaled_matrix	Unscaled data matrix - (num_samples by num_features)

Returns: Shared pointer to a DataScaler

Referenced by GaussianProcess::build(), and PolynomialRegression::build().

◆ solver_factory()

std::shared_ptr< LinearSolverBase > solver_factory ( LinearSolverBase::SOLVER_TYPE type )

Free function to construct LinearSolverBase.

Parameters

[in] type Which solver to construct

Returns: Shared pointer to a LinearSolverBase

Referenced by PolynomialRegression::build().

Variable Documentation

◆ type_name_bimap [1/3]

BimapMetrictypeStr type_name_bimap

static

Initial value:

=
    boost::assign::list_of<BimapMetrictypeStr::relation>
    
    (METRIC_TYPE::SUM_SQUARED, "sum_squared")
    (METRIC_TYPE::MEAN_SQUARED, "mean_squared")
    (METRIC_TYPE::ROOT_MEAN_SQUARED, "root_mean_squared")
    (METRIC_TYPE::SUM_ABS, "sum_abs")
    (METRIC_TYPE::MEAN_ABS, "mean_abs")
    (METRIC_TYPE::MAX_ABS, "max_abs")
    (METRIC_TYPE::ABS_PERCENTAGE_ERROR, "ape")
    (METRIC_TYPE::MEAN_ABS_PERCENTAGE_ERROR, "mape")
    (METRIC_TYPE::R_SQUARED, "rsquared")

Bimap between metric types and names.

Referenced by metric_type(), DataScaler::scaler_type(), and LinearSolverBase::solver_type().

◆ type_name_bimap [2/3]

BimapScalertypeStr type_name_bimap

static

Initial value:

=
    boost::assign::list_of<BimapScalertypeStr::relation>
    
    (SCALER_TYPE::NONE, "none")
    (SCALER_TYPE::STANDARDIZATION, "standardization")
    (SCALER_TYPE::MEAN_NORMALIZATION, "mean normalization")
    (SCALER_TYPE::MINMAX_NORMALIZATION, "min-max normalization")

Bimap between scaler types and names.

◆ type_name_bimap [3/3]

BimapSolvertypeStr type_name_bimap

static

Initial value:

=
    boost::assign::list_of<BimapSolvertypeStr::relation>
    
    (SOLVER_TYPE::CHOLESKY, "cholesky")
    (SOLVER_TYPE::EQ_CONS_LEAST_SQ_REGRESSION, "equality-constrained lsq regression")
    (SOLVER_TYPE::LASSO_REGRESSION, "lasso regression")
    (SOLVER_TYPE::LEAST_ANGLE_REGRESSION, "least angle regression")
    (SOLVER_TYPE::LU, "LU")
    (SOLVER_TYPE::ORTHOG_MATCH_PURSUIT, "orthogonal matching pursuit")
    (SOLVER_TYPE::QR_LEAST_SQ_REGRESSION, "QR lsq regression")
    (SOLVER_TYPE::SVD_LEAST_SQ_REGRESSION, "SVD")

Bimap between solver types and names.

Classes

Typedefs

Enumerations

Functions

Variables

Detailed Description

Function Documentation

◆ error()

◆ matrix_equals() [1/3]

◆ matrix_equals() [2/3]

◆ matrix_equals() [3/3]

◆ relative_allclose()

◆ variance()

◆ populateVectorsFromFile()

◆ populateMatricesFromFile()

◆ n_choose_k()

◆ num_nonzeros()

◆ nonzero()

◆ append_columns()

◆ p_norm()

◆ metric_type()

◆ compute_metric()

◆ scaler_factory()

◆ solver_factory()

Variable Documentation

◆ type_name_bimap [1/3]

◆ type_name_bimap [2/3]

◆ type_name_bimap [3/3]