Performs a line search using the Armijo condition and uses quadratic interpolation to select each new alpha. More...

#include <ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.hpp>

Inheritance diagram for ConstrainedOptPack::DirectLineSearchArmQuad_Strategy:

Public Member Functions
	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, eta)
	Set the Armijo cord test fractional reduction parameter. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, min_frac)
	The minimum fraction that alpha is reduced for each line search iteration. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (value_type, max_frac)
	The maximum fraction that alpha is reduced for each line search iteration. More...

	STANDARD_MEMBER_COMPOSITION_MEMBERS (bool, max_out_iter)
	Deterimine if the line search iterations are maxed out or not. More...

	DirectLineSearchArmQuad_Strategy (int max_iter=20, value_type eta=1.0e-4, value_type min_frac=0.1, value_type max_frac=0.5, bool max_out_iter=false)
	Constructs with default settings. More...

Public Member Functions inherited from ConstrainedOptPack::DirectLineSearch_Strategy
virtual	~DirectLineSearch_Strategy ()

Private Member Functions
void	validate_parameters () const

Private Attributes
int	max_iter_

int	num_iter_

Overridden from DirectLineSearch_Strategy
void	set_max_iter (int max_iter)

int	max_iter () const

int	num_iterations () const

bool	do_line_search (const MeritFuncCalc1D &phi, value_type phi_k, value_type alpha_k, value_type phi_kp1, std::ostream *out)
	Performs the following line search: More...

void	print_algorithm (std::ostream &out, const std::string &leading_str) const

Detailed Description

Performs a line search using the Armijo condition and uses quadratic interpolation to select each new alpha.

Definition at line 53 of file ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.hpp.

Constructor & Destructor Documentation

ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::DirectLineSearchArmQuad_Strategy	(	int	max_iter = `20`,
		value_type	eta = `1.0e-4`,
		value_type	min_frac = `0.1`,
		value_type	max_frac = `0.5`,
		bool	max_out_iter = `false`
	)

Constructs with default settings.

Definition at line 59 of file ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.cpp.

Member Function Documentation

ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		eta
	)

Set the Armijo cord test fractional reduction parameter.

ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		min_frac
	)

The minimum fraction that alpha is reduced for each line search iteration.

ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	value_type	,
		max_frac
	)

The maximum fraction that alpha is reduced for each line search iteration.

ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::STANDARD_MEMBER_COMPOSITION_MEMBERS	(	bool	,
		max_out_iter
	)

Deterimine if the line search iterations are maxed out or not.

This option is really only used for debugging and requires changing the other parameters to make it useful.

void ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::set_max_iter ( int max_iter )

virtual

Implements ConstrainedOptPack::DirectLineSearch_Strategy.

Definition at line 73 of file ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.cpp.

int ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::max_iter ( ) const

virtual

Implements ConstrainedOptPack::DirectLineSearch_Strategy.

Definition at line 78 of file ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.cpp.

int ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::num_iterations ( ) const

virtual

Implements ConstrainedOptPack::DirectLineSearch_Strategy.

Definition at line 83 of file ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.cpp.

bool ConstrainedOptPack::DirectLineSearchArmQuad_Strategy::do_line_search	(	const MeritFuncCalc1D &	phi,
		value_type	phi_k,
		value_type *	alpha_k,
		value_type *	phi_kp1,
		std::ostream *	out
	)

virtual

Performs the following line search:

num_iter = 0;
while( phi.value(alpha_k) > phi_k + eta * alpha_k * phi.deriv() ) 
{
   if(num_iter >= max_iter) return true;
   num_iter = num_iter + 1;
   alpha_k = [ min_frac * alpha_k <= quadradic interpolation for alpha  <= max_frac * alpha_k ];
}
return true;<br>

If the maximum number of iterations is exceeded then false will be returned.

The default values for the adjustable parameters (from D&S A6.3.1) are:
max_iter = 20
eta = 1.0e-4
min_frac = 0.1
max_frac = 0.5

Implements ConstrainedOptPack::DirectLineSearch_Strategy.

Definition at line 88 of file ConstrainedOptPack_DirectLineSearchArmQuad_Strategy.cpp.