This class provides options for a Optimizer. More...

#include <OptimizerOptions.h>

Collaboration diagram for QUESO::OptimizerOptions:

Public Member Functions
Constructor/Destructor methods
	OptimizerOptions ()
	Default constructor. All options have their default values. More...

	OptimizerOptions (const BaseEnvironment env, const char prefix="")
	A constructor that takes the environment for parsing input file options. More...

	OptimizerOptions (const OptimizerOptions &rhs)
	Copy constructor. More...

virtual	~OptimizerOptions ()
	Destructor. More...

Public Attributes
std::string	m_prefix

std::string	m_help
	If this string is non-empty, options are print to the output file. More...

unsigned int	m_maxIterations
	The maximum number of iterations to do for optimization. Default is 100. More...

double	m_tolerance
	The tolerance at which optimization stops. Default is 1e-3. More...

double	m_finiteDifferenceStepSize
	The step size used to compute gradients from finite differencing. Default is 1e-4. More...

std::string	m_solverType
	The optimization algorithm to use. Default is bfgs2. More...

double	m_fstepSize
	The size of the initial trial steps for optimizing without gradients. Default is 0.1. More...

double	m_fdfstepSize
	The size of the first step when optimizing with gradients. Default is 1.0. More...

double	m_lineTolerance
	Accuracy to which to solve line minization to. Default is 0.1. More...

Private Member Functions
void	checkOptions ()

Private Attributes
const BaseEnvironment *	m_env

BoostInputOptionsParser *	m_parser

std::string	m_option_help

std::string	m_option_maxIterations
	Option name for OptimizerOptions::m_maxIterations. Default is m_prefix + "optimizer_maxIterations". More...

std::string	m_option_tolerance
	Option name for OptimizerOptions::m_tolerance. Default is m_prefix + "optimizer_tolerance". More...

std::string	m_option_finiteDifferenceStepSize
	Option name for OptimizerOptions::m_finiteDifferenceStepSize. Default is m_prefix + "optimizer_finiteDifferenceStepSize". More...

std::string	m_option_solverType
	Option name for OptimizerOptions::m_solverType. Default is m_prefix + "optimizer_solverType". More...

std::string	m_option_fstepSize
	Option name for OptimizerOptions::m_fstepSize. Default is m_prefix + "optimizer_fstepSize". More...

std::string	m_option_fdfstepSize
	Option name for OptimizerOptions::m_fdfstepSize. Default is m_prefix + "optimizer_fdfStepSize". More...

std::string	m_option_lineTolerance
	Option name for OptimizerOptions::m_lineTolerance. Default is m_prefix + "optimizer_lineTolerance". More...

Friends
std::ostream &	operator<< (std::ostream &os, const OptimizerOptions &obj)

Detailed Description

This class provides options for a Optimizer.

Definition at line 60 of file OptimizerOptions.h.

Constructor & Destructor Documentation

QUESO::OptimizerOptions::OptimizerOptions ( )

Default constructor. All options have their default values.

Definition at line 32 of file OptimizerOptions.C.

   : m_prefix("ip_"),
     m_help(UQ_OPT_HELP),
     m_maxIterations(UQ_OPT_MAX_ITERATIONS),
     m_tolerance(UQ_OPT_TOLERANCE),
     m_finiteDifferenceStepSize(UQ_OPT_FINITE_DIFFERENCE_STEP_SIZE),
     m_solverType(UQ_OPT_SOLVER_TYPE),
     m_fstepSize(UQ_OPT_FSTEP_SIZE),
     m_fdfstepSize(UQ_OPT_FDFSTEP_SIZE),
     m_lineTolerance(UQ_OPT_LINE_TOLERANCE),
     m_env(NULL),
 #ifndef DISABLE_BOOST_PROGRAM_OPTIONS
     m_parser(NULL),
 #endif  // DISABLE_BOOST_PROGRAM_OPTIONS
     m_option_help(m_prefix + "help"),
     m_option_maxIterations(m_prefix + "maxIterations"),
     m_option_tolerance(m_prefix + "tolerance"),
     m_option_finiteDifferenceStepSize(m_prefix + "finiteDifferenceStepSize"),
     m_option_solverType(m_prefix + "solverType"),
     m_option_fstepSize(m_prefix + "fstepSize"),
     m_option_fdfstepSize(m_prefix + "fdfStepSize"),
     m_option_lineTolerance(m_prefix + "lineTolerance")
 {
 }

QUESO::OptimizerOptions::OptimizerOptions	(	const BaseEnvironment *	env,
		const char *	prefix = `""`
	)

A constructor that takes the environment for parsing input file options.

Definition at line 57 of file OptimizerOptions.C.

References checkOptions(), QUESO::BoostInputOptionsParser::getOption(), QUESO::BaseEnvironment::input(), m_env, m_fdfstepSize, m_finiteDifferenceStepSize, m_fstepSize, m_help, m_lineTolerance, m_maxIterations, m_option_fdfstepSize, m_option_finiteDifferenceStepSize, m_option_fstepSize, m_option_help, m_option_lineTolerance, m_option_maxIterations, m_option_solverType, m_option_tolerance, m_parser, m_solverType, m_tolerance, QUESO::BoostInputOptionsParser::registerOption(), QUESO::BoostInputOptionsParser::scanInputFile(), UQ_OPT_FDFSTEP_SIZE, UQ_OPT_FINITE_DIFFERENCE_STEP_SIZE, UQ_OPT_FSTEP_SIZE, UQ_OPT_HELP, UQ_OPT_LINE_TOLERANCE, UQ_OPT_MAX_ITERATIONS, UQ_OPT_SOLVER_TYPE, and UQ_OPT_TOLERANCE.

   : m_prefix((std::string)(prefix) + "optimizer_"),
     m_help(UQ_OPT_HELP),
     m_maxIterations(UQ_OPT_MAX_ITERATIONS),
     m_tolerance(UQ_OPT_TOLERANCE),
     m_finiteDifferenceStepSize(UQ_OPT_FINITE_DIFFERENCE_STEP_SIZE),
     m_solverType(UQ_OPT_SOLVER_TYPE),
     m_fstepSize(UQ_OPT_FSTEP_SIZE),
     m_fdfstepSize(UQ_OPT_FDFSTEP_SIZE),
     m_lineTolerance(UQ_OPT_LINE_TOLERANCE),
     m_env(env),
 #ifndef DISABLE_BOOST_PROGRAM_OPTIONS
     m_parser(new BoostInputOptionsParser(env->optionsInputFileName())),
 #endif  // DISABLE_BOOST_PROGRAM_OPTIONS
     m_option_help(m_prefix + "help"),
     m_option_maxIterations(m_prefix + "maxIterations"),
     m_option_tolerance(m_prefix + "tolerance"),
     m_option_finiteDifferenceStepSize(m_prefix + "finiteDifferenceStepSize"),
     m_option_solverType(m_prefix + "solverType"),
     m_option_fstepSize(m_prefix + "fstepSize"),
     m_option_fdfstepSize(m_prefix + "fdfStepSize"),
     m_option_lineTolerance(m_prefix + "lineTolerance")
 {
 #ifndef DISABLE_BOOST_PROGRAM_OPTIONS
   m_parser->registerOption<std::string>(m_option_help, UQ_OPT_HELP,
       "produce help message for statistical inverse problem");
 
   m_parser->registerOption<unsigned int>(m_option_maxIterations,
       UQ_OPT_MAX_ITERATIONS,
       "max number of optimizer iterations to do");
   m_parser->registerOption<double>(m_option_tolerance, UQ_OPT_TOLERANCE,
       "optimize until gradient is less than tolerance");
   m_parser->registerOption<double>(m_option_finiteDifferenceStepSize,
       UQ_OPT_FINITE_DIFFERENCE_STEP_SIZE,
       "if no deriv is given, do finite difference with this step size");
   m_parser->registerOption<std::string>(m_option_solverType, UQ_OPT_SOLVER_TYPE,
       "which optimisation algorithm to use");
   m_parser->registerOption<double>(m_option_fstepSize, UQ_OPT_FSTEP_SIZE,
       "sets the step size used in gradient-free solvers");
   m_parser->registerOption<double>(m_option_fdfstepSize, UQ_OPT_FDFSTEP_SIZE,
       "sets the step size used in gradient-based solvers");
   m_parser->registerOption<double>(m_option_lineTolerance, UQ_OPT_LINE_TOLERANCE,
       "sets the line minimisation tolerance");
 
   m_parser->scanInputFile();
 
   m_parser->getOption<std::string>(m_option_help, m_help);
   m_parser->getOption<unsigned int>(m_option_maxIterations, m_maxIterations);
   m_parser->getOption<double>(m_option_tolerance, m_tolerance);
   m_parser->getOption<double>(m_option_finiteDifferenceStepSize,
       m_finiteDifferenceStepSize);
   m_parser->getOption<std::string>(m_option_solverType, m_solverType);
   m_parser->getOption<double>(m_option_fstepSize, m_fstepSize);
   m_parser->getOption<double>(m_option_fdfstepSize, m_fdfstepSize);
   m_parser->getOption<double>(m_option_lineTolerance, m_lineTolerance);
 #else
   m_help = m_env->input()(m_option_help, UQ_OPT_HELP);
   m_maxIterations = m_env->input()(m_option_maxIterations, UQ_OPT_MAX_ITERATIONS);
   m_tolerance = m_env->input()(m_option_tolerance, UQ_OPT_TOLERANCE);
   m_finiteDifferenceStepSize = m_env->input()(m_option_finiteDifferenceStepSize, UQ_OPT_FINITE_DIFFERENCE_STEP_SIZE);
   m_solverType = m_env->input()(m_option_solverType, UQ_OPT_SOLVER_TYPE);
   m_fstepSize = m_env->input()(m_option_fstepSize, UQ_OPT_FSTEP_SIZE);
   m_fdfstepSize = m_env->input()(m_option_fdfstepSize, UQ_OPT_FDFSTEP_SIZE);
   m_lineTolerance = m_env->input()(m_option_lineTolerance, UQ_OPT_LINE_TOLERANCE);
 #endif  // DISABLE_BOOST_PROGRAM_OPTIONS
 
   checkOptions();
 }

QUESO::OptimizerOptions::OptimizerOptions ( const OptimizerOptions & rhs )

Copy constructor.

Definition at line 127 of file OptimizerOptions.C.

   :
     m_prefix(rhs.m_prefix),
     m_help(rhs.m_help),
     m_maxIterations(rhs.m_maxIterations),
     m_tolerance(rhs.m_tolerance),
     m_finiteDifferenceStepSize(rhs.m_finiteDifferenceStepSize),
     m_solverType(rhs.m_solverType),
     m_fstepSize(rhs.m_fstepSize),
     m_fdfstepSize(rhs.m_fdfstepSize),
     m_lineTolerance(rhs.m_lineTolerance),
     m_parser(rhs.m_parser),  // We'll never touch the input file in a copied object
     m_option_help(rhs.m_option_help),
     m_option_maxIterations(rhs.m_option_maxIterations),
     m_option_tolerance(rhs.m_option_tolerance),
     m_option_finiteDifferenceStepSize(rhs.m_option_finiteDifferenceStepSize),
     m_option_solverType(rhs.m_option_solverType),
     m_option_fstepSize(rhs.m_option_fstepSize),
     m_option_fdfstepSize(rhs.m_option_fdfstepSize),
     m_option_lineTolerance(rhs.m_option_lineTolerance)
 {
 }

QUESO::OptimizerOptions::~OptimizerOptions ( )

virtual

Destructor.

Definition at line 150 of file OptimizerOptions.C.

151 {

152 }

Member Function Documentation

void QUESO::OptimizerOptions::checkOptions ( )

private

Definition at line 155 of file OptimizerOptions.C.

References m_fdfstepSize, m_finiteDifferenceStepSize, m_fstepSize, m_lineTolerance, m_maxIterations, m_tolerance, and queso_require_greater_msg.

Referenced by OptimizerOptions().

 {
   queso_require_greater_msg(m_tolerance, 0.0, "optimizer tolerance must be > 0");
   queso_require_greater_msg(m_finiteDifferenceStepSize, 0.0, "finite difference step must be > 0");
   queso_require_greater_msg(m_maxIterations, 0, "max iterations must be > 0");
   queso_require_greater_msg(m_fstepSize, 0.0, "fstepSize must be > 0");
   queso_require_greater_msg(m_fdfstepSize, 0.0, "fdfstepSize must be > 0");
   queso_require_greater_msg(m_lineTolerance, 0.0, "line tolerance must be > 0");
 }

Friends And Related Function Documentation

std::ostream& operator<<	(	std::ostream &	os,
		const OptimizerOptions &	obj
	)

friend

Definition at line 166 of file OptimizerOptions.C.

 {
   os << "\n" << obj.m_option_maxIterations << " = " << obj.m_maxIterations
      << "\n" << obj.m_option_tolerance << " = " << obj.m_tolerance;
   os << "\n" << obj.m_option_finiteDifferenceStepSize << " = "
              << obj.m_finiteDifferenceStepSize;
   os << "\n" << obj.m_option_solverType << " = " << obj.m_solverType;
   os << "\n" << obj.m_option_fstepSize << " = " << obj.m_fstepSize;
   os << "\n" << obj.m_option_fdfstepSize << " = " << obj.m_fdfstepSize;
   os << "\n" << obj.m_option_lineTolerance << " = " << obj.m_lineTolerance;
   os << std::endl;
   return os;
 }

Member Data Documentation

const BaseEnvironment* QUESO::OptimizerOptions::m_env

private

Definition at line 136 of file OptimizerOptions.h.

Referenced by OptimizerOptions().

double QUESO::OptimizerOptions::m_fdfstepSize

The size of the first step when optimizing with gradients. Default is 1.0.

Definition at line 126 of file OptimizerOptions.h.

Referenced by checkOptions(), QUESO::operator<<(), and OptimizerOptions().

double QUESO::OptimizerOptions::m_finiteDifferenceStepSize

The step size used to compute gradients from finite differencing. Default is 1e-4.

This is only done when analytical gradients are not provided (either by the user or by QUESO).

Definition at line 101 of file OptimizerOptions.h.

Referenced by checkOptions(), QUESO::operator<<(), and OptimizerOptions().

double QUESO::OptimizerOptions::m_fstepSize

The size of the initial trial steps for optimizing without gradients. Default is 0.1.

The initial trial steps is actually a vector of size equal to the number of parameters, and this vector is set to m_fstepSize in each component.

Definition at line 123 of file OptimizerOptions.h.

Referenced by checkOptions(), QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_help

If this string is non-empty, options are print to the output file.

Definition at line 81 of file OptimizerOptions.h.

Referenced by OptimizerOptions().

double QUESO::OptimizerOptions::m_lineTolerance

Accuracy to which to solve line minization to. Default is 0.1.

The exact meaning of this parameter depends on the method used. It's not used for algorithms that don't use derivatives.

Definition at line 133 of file OptimizerOptions.h.

Referenced by checkOptions(), QUESO::operator<<(), and OptimizerOptions().

unsigned int QUESO::OptimizerOptions::m_maxIterations

The maximum number of iterations to do for optimization. Default is 100.

Definition at line 84 of file OptimizerOptions.h.

Referenced by checkOptions(), QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_fdfstepSize

private

Option name for OptimizerOptions::m_fdfstepSize. Default is m_prefix + "optimizer_fdfStepSize".

Definition at line 154 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_finiteDifferenceStepSize

private

Option name for OptimizerOptions::m_finiteDifferenceStepSize. Default is m_prefix + "optimizer_finiteDifferenceStepSize".

Definition at line 148 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_fstepSize

private

Option name for OptimizerOptions::m_fstepSize. Default is m_prefix + "optimizer_fstepSize".

Definition at line 152 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_help

private

Definition at line 141 of file OptimizerOptions.h.

Referenced by OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_lineTolerance

private

Option name for OptimizerOptions::m_lineTolerance. Default is m_prefix + "optimizer_lineTolerance".

Definition at line 156 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_maxIterations

private

Option name for OptimizerOptions::m_maxIterations. Default is m_prefix + "optimizer_maxIterations".

Definition at line 144 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_solverType

private

Option name for OptimizerOptions::m_solverType. Default is m_prefix + "optimizer_solverType".

Definition at line 150 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

std::string QUESO::OptimizerOptions::m_option_tolerance

private

Option name for OptimizerOptions::m_tolerance. Default is m_prefix + "optimizer_tolerance".

Definition at line 146 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

BoostInputOptionsParser* QUESO::OptimizerOptions::m_parser

private

Definition at line 138 of file OptimizerOptions.h.

Referenced by OptimizerOptions().

std::string QUESO::OptimizerOptions::m_prefix

Definition at line 78 of file OptimizerOptions.h.

std::string QUESO::OptimizerOptions::m_solverType

The optimization algorithm to use. Default is bfgs2.

Choices are:

fletcher_reeves_cg
polak_ribiere_cg
bfgs
bfgs2
steepest_descent
nelder_mead
nelder_mead2
nelder_mead2_rand

Definition at line 115 of file OptimizerOptions.h.

Referenced by QUESO::operator<<(), and OptimizerOptions().

double QUESO::OptimizerOptions::m_tolerance

The tolerance at which optimization stops. Default is 1e-3.

The tolerance is measured in terms of the 2-norm of the gradient of the objective function for optimizers that use derivatives.

For optimizers that do not use derivatives, the tolerance is measured in terms of the size of the size of the simplex for the simplex optimizers.

Definition at line 94 of file OptimizerOptions.h.

Referenced by checkOptions(), QUESO::operator<<(), and OptimizerOptions().

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

src/core/inc/OptimizerOptions.h
src/core/src/OptimizerOptions.C

Public Member Functions

Public Attributes

Private Member Functions

Private Attributes

Friends

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Friends And Related Function Documentation

Member Data Documentation