math_GlobOptMin Class Reference

This class represents Evtushenko's algorithm of global optimization based on non-uniform mesh. Article: Yu. Evtushenko. Numerical methods for finding global extreme (case of a non-uniform mesh). U.S.S.R. Comput. Maths. Math. Phys., Vol. 11, N 6, pp. 38-54. More...

#include <math_GlobOptMin.hxx>

Public Member Functions
	math_GlobOptMin (math_MultipleVarFunction *theFunc, const math_Vector &theLowerBorder, const math_Vector &theUpperBorder, const double theC=9, const double theDiscretizationTol=1.0e-2, const double theSameTol=1.0e-7)
	Constructor. Perform method is not called from it.
void	SetGlobalParams (math_MultipleVarFunction *theFunc, const math_Vector &theLowerBorder, const math_Vector &theUpperBorder, const double theC=9, const double theDiscretizationTol=1.0e-2, const double theSameTol=1.0e-7)
void	SetLocalParams (const math_Vector &theLocalA, const math_Vector &theLocalB)
	Method to reduce bounding box. Perform will use this box.
void	SetTol (const double theDiscretizationTol, const double theSameTol)
	Method to set tolerances.
void	GetTol (double &theDiscretizationTol, double &theSameTol)
	Method to get tolerances.
void	Perform (const bool isFindSingleSolution=false)
void	Points (const int theIndex, math_Vector &theSol)
	Return solution theIndex, 1 <= theIndex <= NbExtrema.
void	SetContinuity (const int theCont)
	Set / Get continuity of local borders splits (0 ~ C0, 1 ~ C1, 2 ~ C2).
int	GetContinuity () const
void	SetFunctionalMinimalValue (const double theMinimalValue)
	Set / Get functional minimal value.
double	GetFunctionalMinimalValue () const
void	SetLipConstState (const bool theFlag)
	Set / Get Lipchitz constant modification state. True means that the constant is locked and unlocked otherwise.
bool	GetLipConstState () const
bool	isDone () const
	Return computation state of the algorithm.
double	GetF () const
	Get best functional value.
int	NbExtrema () const
	Return count of global extremas.

Detailed Description

This class represents Evtushenko's algorithm of global optimization based on non-uniform mesh. Article: Yu. Evtushenko. Numerical methods for finding global extreme (case of a non-uniform mesh). U.S.S.R. Comput. Maths. Math. Phys., Vol. 11, N 6, pp. 38-54.

This method performs search on non-uniform mesh. The search space is a box in R^n space. The default behavior is to find all minimums in that box. Computation of maximums is not supported.

The search box can be split into smaller boxes by discontinuity criteria. This functionality is covered by SetGlobalParams and SetLocalParams API.

It is possible to set continuity of the local boxes. Such option can forcibly change local extrema search. In other words if theFunc can be casted to the function with Hessian but, continuity is set to 1 Gradient based local optimization method will be used, not Hessian based method. This functionality is covered by SetContinuity and GetContinuity API.

It is possible to freeze Lipschitz const to avoid internal modifications on it. This functionality is covered by SetLipConstState and GetLipConstState API.

It is possible to perform single solution search. This functionality is covered by first parameter in Perform method.

It is possible to set / get minimal value of the functional. It works well together with single solution search. This functionality is covered by SetFunctionalMinimalValue and GetFunctionalMinimalValue API.

Constructor & Destructor Documentation

math_GlobOptMin()

math_GlobOptMin::math_GlobOptMin	(	math_MultipleVarFunction *	theFunc,
		const math_Vector &	theLowerBorder,
		const math_Vector &	theUpperBorder,
		const double	theC = 9,
		const double	theDiscretizationTol = 1.0e-2,
		const double	theSameTol = 1.0e-7 )

Constructor. Perform method is not called from it.

Parameters

theFunc	- objective functional.
theLowerBorder	- lower corner of the search box.
theUpperBorder	- upper corner of the search box.
theC	- Lipschitz constant.
theDiscretizationTol	- parameter space discretization tolerance.
theSameTol	- functional value space indifference tolerance.

Member Function Documentation

GetContinuity()

int math_GlobOptMin::GetContinuity ( ) const

inline

GetF()

double math_GlobOptMin::GetF ( ) const

inline

Get best functional value.

GetFunctionalMinimalValue()

double math_GlobOptMin::GetFunctionalMinimalValue ( ) const

inline

GetLipConstState()

bool math_GlobOptMin::GetLipConstState ( ) const

inline

GetTol()

void math_GlobOptMin::GetTol	(	double &	theDiscretizationTol,
		double &	theSameTol )

Method to get tolerances.

Parameters

theDiscretizationTol	- parameter space discretization tolerance.
theSameTol	- functional value space indifference tolerance.

isDone()

bool math_GlobOptMin::isDone ( ) const

inline

Return computation state of the algorithm.

NbExtrema()

int math_GlobOptMin::NbExtrema ( ) const

inline

Return count of global extremas.

Perform()

void math_GlobOptMin::Perform

(

const bool

isFindSingleSolution = false

)

Parameters

isFindSingleSolution

- defines whether to find single solution or all solutions.

Points()

void math_GlobOptMin::Points	(	const int	theIndex,
		math_Vector &	theSol )

Return solution theIndex, 1 <= theIndex <= NbExtrema.

SetContinuity()

void math_GlobOptMin::SetContinuity ( const int theCont )

inline

Set / Get continuity of local borders splits (0 ~ C0, 1 ~ C1, 2 ~ C2).

SetFunctionalMinimalValue()

void math_GlobOptMin::SetFunctionalMinimalValue ( const double theMinimalValue )

inline

Set / Get functional minimal value.

SetGlobalParams()

void math_GlobOptMin::SetGlobalParams	(	math_MultipleVarFunction *	theFunc,
		const math_Vector &	theLowerBorder,
		const math_Vector &	theUpperBorder,
		const double	theC = 9,
		const double	theDiscretizationTol = 1.0e-2,
		const double	theSameTol = 1.0e-7 )

Parameters

theFunc	- objective functional.
theLowerBorder	- lower corner of the search box.
theUpperBorder	- upper corner of the search box.
theC	- Lipschitz constant.
theDiscretizationTol	- parameter space discretization tolerance.
theSameTol	- functional value space indifference tolerance.

SetLipConstState()

void math_GlobOptMin::SetLipConstState ( const bool theFlag )

inline

Set / Get Lipchitz constant modification state. True means that the constant is locked and unlocked otherwise.

SetLocalParams()

void math_GlobOptMin::SetLocalParams	(	const math_Vector &	theLocalA,
		const math_Vector &	theLocalB )

Method to reduce bounding box. Perform will use this box.

Parameters

theLocalA	- lower corner of the local box.
theLocalB	- upper corner of the local box.

SetTol()

void math_GlobOptMin::SetTol	(	const double	theDiscretizationTol,
		const double	theSameTol )

Method to set tolerances.

Parameters

theDiscretizationTol	- parameter space discretization tolerance.
theSameTol	- functional value space indifference tolerance.

---

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

• math_GlobOptMin.hxx