NLopt optimization. More...

#include <NLoptSolver.hxx>

Public Member Functions
	NLoptSolver ()
	Default constructor.
	~NLoptSolver ()
	Destructor.
void	Initialize (int Nparameter, string algorithm, double parameter_tolerance=1.e-6, double cost_function_tolerance=1.e-6, int Niteration_max=-1)
	Initializations.
void	SetLowerBound (const Vector< double > &)
	Sets lower bounds on the parameters.
void	SetUpperBound (const Vector< double > &)
	Sets upper bounds on the parameters.
void	SetParameterTolerance (double)
	Sets the relative tolerance on the parameters.
void	SetCostFunctionTolerance (double)
	Sets the relative tolerance on the cost function.
void	SetNiterationMax (int)
	Sets the maximum number of cost function evaluations.
void	GetParameterTolerance (double &) const
	Gets the relative tolerance on the parameters.
void	GetCostFunctionTolerance (double &) const
	Gets the relative tolerance on the cost function.
void	GetNiterationMax (int &) const
	Gets the maximum number of cost function evaluations.
void	SetParameter (const Vector< double > &parameter)
	Sets the parameters.
void	GetParameter (Vector< double > &parameter) const
	Gets the parameters.
void	Optimize (cost_ptr cost, void *argument)
	Optimization.
double	GetCost () const
	Returns the value of the cost function.
Protected Types
typedef double(*	cost_ptr )(const Vector< double > &, Vector< double > &, void *)
Protected Attributes
nlopt::SeldonOpt	opt_
	NLopt optimization solver.
nlopt::algorithm	algorithm_
	Optimization algorithm.
double	parameter_tolerance_
	Relative tolerance on the optimization parameters.
double	cost_function_tolerance_
	Relative tolerance on the cost function.
int	Niteration_max_
	Maximum number of function evaluations. It is ignored if it is non-positive.
Vector< double >	parameter_
	The vector that stores parameters values. Before optimization, stores the initial parameter vector; after optimization, it returns the optimized parameters.
Vector< double >	gradient_
	The vector that stores gradient values. Before optimization, unspecified; after optimization, it returns the gradient vector for optimized parameters.
double	cost_
	The value of cost function for given parameter values.

Detailed Description

NLopt optimization.

Definition at line 38 of file NLoptSolver.hxx.

Member Function Documentation

double Seldon::NLoptSolver::GetCost ( ) const

Returns the value of the cost function.

This method should be called after the optimization, and it returns the value of the cost function associated with the optimized parameters.

Returns:: the value of the cost function.

Definition at line 314 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::GetCostFunctionTolerance ( double & tolerance ) const

Gets the relative tolerance on the cost function.

Parameters:

[out] tolerance relative tolerance on the cost function . When the variation of the cost function, after one step of the algorithm, has changed by less than tolerance multiplied by the value of the cost function, the optimization is stopped.

Definition at line 238 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::GetNiterationMax ( int & Niteration_max ) const

Gets the maximum number of cost function evaluations.

Parameters:

[out] Niteration_max maximum number of cost function evaluations.

Definition at line 248 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::GetParameter ( Vector< double > & parameter ) const

Gets the parameters.

Parameters:

[out] parameter the parameters vector.

Definition at line 269 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::GetParameterTolerance ( double & tolerance ) const

Gets the relative tolerance on the parameters.

Parameters:

[out] tolerance relative tolerance on the parameters . When the variation of every parameter, after one step of the algorithm, has changed by less than tolerance multiplied by the value of the parameter, the optimization is stopped.

Definition at line 225 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::Initialize	(	int	Nparameter,
		string	algorithm,
		double	parameter_tolerance = `1.e-6`,
		double	cost_function_tolerance = `1.e-6`,
		int	Niteration_max = `-1`
	)

Initializations.

Parameters:

`[in]`	Nparameter	total number of parameters to be optimized.
`[in]`	algorithm	name of the optimization algorithm, one of: GN_DIRECT, GN_DIRECT_L, GN_DIRECT_L_RAND, GN_DIRECT_NOSCAL, GN_DIRECT_L_NOSCAL, GN_DIRECT_L_RAND_NOSCAL, GN_ORIG_DIRECT, GN_ORIG_DIRECT_L, GD_STOGO, GD_STOGO_RAND, LD_LBFGS_NOCEDAL, LD_LBFGS, LN_PRAXIS, LD_VAR1, LD_VAR2, LD_TNEWTON, LD_TNEWTON_RESTART, LD_TNEWTON_PRECOND, LD_TNEWTON_PRECOND_RESTART, GN_CRS2_LM, GN_MLSL, GD_MLSL, GN_MLSL_LDS, GD_MLSL_LDS, LD_MMA, LN_COBYLA, LN_NEWUOA, LN_NEWUOA_BOUND, LN_NELDERMEAD, LN_SBPLX, LN_AUGLAG, LD_AUGLAG, LN_AUGLAG_EQ, LD_AUGLAG_EQ, LN_BOBYQA, GN_ISRES, AUGLAG, AUGLAG_EQ, G_MLSL, G_MLSL_LDS, LD_SLSQP, NUM_ALGORITHMS.
`[in]`	parameter_tolerance	relative tolerance on the parameters. When the variation of the parameters, after one step of the algorithm, has changed by less than parameter_tolerance multiplied by the value of the parameters, the optimization is stopped. If you do not want to use a particular tolerance termination, you can just set that tolerance to zero and it will be ignored.
`[in]`	cost_function_tolerance	relative tolerance on the cost function. When the variation of the cost function, after one step of the algorithm, has changed by less than cost_function_tolerance multiplied by the value of the cost function, the optimization is stopped. If you do not want to use a particular tolerance termination, you can just set that tolerance to zero and it will be ignored.
`[in]`	Niteration_max	maximum number of cost function evaluations. It is ignored if it is non-positive.

Definition at line 69 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::Optimize	(	cost_ptr	cost,
		void *	argument
	)

Optimization.

Parameters:

`[in]`	cost	pointer to the cost function. This function takes as first argument a 'const Vector<double>&' of parameters. The second argument of the function is a 'Vector<double>&' which must be, on exit, the gradient of the cost function. NLopt will allocate it before the call. Note that, in case a derivative-free algorithm is used, this gradient vector is empty, and the cost function is not supposed to compute it (it is thus recommended to test the length of the vector in the cost function). The third argument of the function is argument, provided as 'void *'. The cost function returns the cost value in 'double'.
`[in]`	argument	third argument of the cost function. This argument is passed to the cost function after the parameters vector and the gradient.

Definition at line 290 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::SetCostFunctionTolerance ( double tolerance )

Sets the relative tolerance on the cost function.

Parameters:

[in] tolerance relative tolerance on the cost function. When the variation of the cost function, after one step of the algorithm, has changed by less than tolerance multiplied by the value of the cost function, the optimization is stopped. If you do not want to use a particular tolerance termination, you can just set that tolerance to zero and it will be ignored.

Definition at line 201 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::SetLowerBound ( const Vector< double > & lower_bound )

Sets lower bounds on the parameters.

Parameters:

[in] lower_bound the lower bound vector.

Definition at line 159 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::SetNiterationMax ( int Niteration_max )

Sets the maximum number of cost function evaluations.

Parameters:

[in] Niteration_max maximum number of cost function evaluations. It is ignored if it is non-positive.

Definition at line 212 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::SetParameter ( const Vector< double > & parameter )

Sets the parameters.

Parameters:

[in] parameter the parameters vector.

Definition at line 258 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::SetParameterTolerance ( double tolerance )

Sets the relative tolerance on the parameters.

Parameters:

[in] tolerance relative tolerance on the parameters. When the variation of every parameter, after one step of the algorithm, has changed by less than tolerance multiplied by the value of the parameter, the optimization is stopped. If you do not want to use a particular tolerance termination, you can just set that tolerance to zero and it will be ignored.

Definition at line 186 of file NLoptSolver.cxx.

void Seldon::NLoptSolver::SetUpperBound ( const Vector< double > & upper_bound )

Sets upper bounds on the parameters.

Parameters:

[in] upper_bound the lower bound vector.

Definition at line 170 of file NLoptSolver.cxx.

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

/home/vivien/public_html/.src_seldon/computation/optimization/NLoptSolver.hxx
/home/vivien/public_html/.src_seldon/computation/optimization/NLoptSolver.cxx

Seldon::NLoptSolver Class Reference

Public Member Functions

Protected Types

Protected Attributes

Detailed Description

Member Function Documentation