EndCriteria Class Reference

Criteria to end optimization process: More...

#include <ql/math/optimization/endcriteria.hpp>

Public Types
enum	Type {   None , MaxIterations , StationaryPoint , StationaryFunctionValue ,   StationaryFunctionAccuracy , ZeroGradientNorm , FunctionEpsilonTooSmall , Unknown }

Public Member Functions
	EndCriteria (Size maxIterations, Size maxStationaryStateIterations, Real rootEpsilon, Real functionEpsilon, Real gradientNormEpsilon)
	Initialization constructor.
Size	maxIterations () const
Size	maxStationaryStateIterations () const
Real	rootEpsilon () const
Real	functionEpsilon () const
Real	gradientNormEpsilon () const
bool	operator() (Size iteration, Size &statState, bool positiveOptimization, Real fold, Real normgold, Real fnew, Real normgnew, EndCriteria::Type &ecType) const
bool	checkMaxIterations (Size iteration, EndCriteria::Type &ecType) const
bool	checkStationaryPoint (Real xOld, Real xNew, Size &statStateIterations, EndCriteria::Type &ecType) const
bool	checkStationaryFunctionValue (Real fxOld, Real fxNew, Size &statStateIterations, EndCriteria::Type &ecType) const
bool	checkStationaryFunctionAccuracy (Real f, bool positiveOptimization, EndCriteria::Type &ecType) const
bool	checkZeroGradientNorm (Real gNorm, EndCriteria::Type &ecType) const

Static Public Member Functions
static bool	succeeded (EndCriteria::Type ecType)

Protected Attributes
Size	maxIterations_
	Maximum number of iterations.
Size	maxStationaryStateIterations_
	Maximun number of iterations in stationary state.
Real	rootEpsilon_
	root, function and gradient epsilons
Real	functionEpsilon_
Real	gradientNormEpsilon_

Detailed Description

Criteria to end optimization process:

• maximum number of iterations AND minimum number of iterations around stationary point
• x (independent variable) stationary point
• y=f(x) (dependent variable) stationary point
• stationary gradient

Examples

BermudanSwaption.cpp, Gaussian1dModels.cpp, and GlobalOptimizer.cpp.

Member Function Documentation

operator()()

bool operator()	(	Size	iteration,
		Size &	statState,
		bool	positiveOptimization,
		Real	fold,
		Real	normgold,
		Real	fnew,
		Real	normgnew,
		EndCriteria::Type &	ecType ) const

Test if the number of iterations is not too big and if a minimum point is not reached

checkMaxIterations()

bool checkMaxIterations	(	Size	iteration,
		EndCriteria::Type &	ecType ) const

Test if the number of iteration is below MaxIterations

checkStationaryPoint()

bool checkStationaryPoint	(	Real	xOld,
		Real	xNew,
		Size &	statStateIterations,
		EndCriteria::Type &	ecType ) const

Test if the root variation is below rootEpsilon

checkStationaryFunctionValue()

bool checkStationaryFunctionValue	(	Real	fxOld,
		Real	fxNew,
		Size &	statStateIterations,
		EndCriteria::Type &	ecType ) const

Test if the function variation is below functionEpsilon

checkStationaryFunctionAccuracy()

bool checkStationaryFunctionAccuracy	(	Real	f,
		bool	positiveOptimization,
		EndCriteria::Type &	ecType ) const

Test if the function value is below functionEpsilon

checkZeroGradientNorm()

bool checkZeroGradientNorm	(	Real	gNorm,
		EndCriteria::Type &	ecType ) const

Test if the gradient norm variation is below gradientNormEpsilon

Test if the gradient norm value is below gradientNormEpsilon