OPTPP::OptNewtonLike Class Reference

OptNewtonLike is the base class for Newton Methods. More...

Inheritance diagram for OPTPP::OptNewtonLike:

Collaboration diagram for OPTPP::OptNewtonLike:

Public Member Functions
virtual void	acceptStep (int k, int step_type)
int	checkAnalyticFDGrad ()
	Compare the analytic gradient with the finite difference gradient.
virtual int	checkConvg ()
	Check to see if algorithm satisfies the convergence criterion.
virtual int	checkDeriv ()
	Compare the analytic gradient with the finite difference gradient.
virtual NEWMAT::ColumnVector	computeSearch (NEWMAT::SymmetricMatrix &H)
	Compute Newton direction.
virtual int	computeStep (NEWMAT::ColumnVector sk)
	Compute the step length along the Newton direction.
DerivOption	getDerivOption () const
	Set the type of finite difference routine.
int	getFevals () const
int	getGevals () const
real	getGradMult () const
NEWMAT::SymmetricMatrix	getHessian () const
int	getSearchSize () const
SearchStrategy	getSearchStrategy () const
	Set globalization strategy for optimization algorithms.
real	getTRSize () const
bool	getWarmStart () const
virtual void	initHessian ()
	Compute the Hessian or its approximation at the initial point.
virtual void	initOpt ()
	Initialize algorithmic parameters.
virtual double	initTrustRegionSize () const
	Initialize the size of the trust-region.
virtual void	optimize ()
	Invoke Newton's method on an unconstrained problem.
	OptNewtonLike (int n, TOLS t)
	OptNewtonLike (int n, UPDATEFCN u)
	OptNewtonLike (int n)
	OptNewtonLike ()
	Default Constructor.
void	printStatus (char *)
	Print status of unconstrained Newton's method.
virtual void	readOptInput ()
	Read user-specified input options from a file.
virtual void	reset ()
	Reset parameters.
void	setDerivOption (DerivOption d)
void	setGradMult (real tau)
	Set gradient multiplier which is used to compute trust-region radius.
void	setHessian (NEWMAT::SymmetricMatrix &H)
	Store the current Hessian matrix.
void	setSearchSize (int sss)
	Set number of points in search scheme for trust-pds search strategy.
void	setSearchStrategy (SearchStrategy s)
void	setTRSize (real delta)
	Set trust-region radius.
virtual NEWMAT::SymmetricMatrix	updateH (NEWMAT::SymmetricMatrix &H, int k)=0
	Compute the Hessian of the objective function or its approximation at the current point.
virtual void	updateModel (int k, int ndim, NEWMAT::ColumnVector x)
void	UseWarmStart (NEWMAT::SymmetricMatrix &H)
virtual	~OptNewtonLike ()
	Destructor.
Protected Member Functions
void	defaultAcceptStep (int, int)
	Provide default implementation of AcceptStep.
NEWMAT::ColumnVector	defaultComputeSearch (NEWMAT::SymmetricMatrix &)
virtual NLP1 *	nlprob () const =0
	Returns pointer to an NLP1 object.
Protected Attributes
DerivOption	finitediff
	User-specified derivative option.
NEWMAT::ColumnVector	gprev
	Gradient at the prev. iteration.
int	grad_evals
	Number of gradient evaluations.
real	gradMult
	Gradient multiplier to compute TR_size.
NEWMAT::SymmetricMatrix	Hessian
	Current Hessian.
int	searchSize
	Search pattern size for TRPDS.
SearchStrategy	strategy
	User-specified globalization Strategy.
real	TR_size
	Trust region radius.
bool	WarmStart
Friends
int	trustpds (NLP1 , ostream , NEWMAT::SymmetricMatrix &, NEWMAT::ColumnVector &, NEWMAT::ColumnVector &, real &, real &, real stpmax, real stpmin, int)
int	trustregion (NLP1 , ostream , NEWMAT::SymmetricMatrix &, NEWMAT::ColumnVector &, NEWMAT::ColumnVector &, real &, real &, real stpmax, real stpmin)

Detailed Description

OptNewtonLike is the base class for Newton Methods.

OptNewtonLike provides common data and functionality for OptFDNewton, OptQNewton, and OptNewton methods.

Note:: Modified by P.J. Williams, pwillia@sandia.gov

Date:: Last modified 03/2007

Definition at line 25 of file OptNewtonLike.h.

Constructor & Destructor Documentation

OPTPP::OptNewtonLike::OptNewtonLike ( ) [inline]

Default Constructor.

See also:: OptNewtonLike(int n); OptNewtonLike(int n, UPDATEFCN u); OptNewtonLike(int n, TOLS t)

Definition at line 48 of file OptNewtonLike.h.

OPTPP::OptNewtonLike::OptNewtonLike ( int n ) [inline]

Parameters:

n

an integer argument.

Definition at line 53 of file OptNewtonLike.h.

OPTPP::OptNewtonLike::OptNewtonLike	(	int	n,
		UPDATEFCN	u
	)			`[inline]`

Parameters:

	n	an integer argument.
	u	a function pointer.

Definition at line 62 of file OptNewtonLike.h.

References OPTPP::OptimizeClass::update_fcn.

OPTPP::OptNewtonLike::OptNewtonLike	(	int	n,
		TOLS	t
	)			`[inline]`

Parameters:

	n	an integer argument.
	t	tolerance class reference.

Definition at line 70 of file OptNewtonLike.h.

virtual OPTPP::OptNewtonLike::~OptNewtonLike ( ) [inline, virtual]

Destructor.

Definition at line 78 of file OptNewtonLike.h.

Member Function Documentation

virtual void OPTPP::OptNewtonLike::acceptStep	(	int	,
		int
	)			`[inline, virtual]`

Note:: Pure virtual functions; Each derived class must define these functions for themselves

Implements OPTPP::OptimizeClass.

Definition at line 87 of file OptNewtonLike.h.

References defaultAcceptStep().

int OPTPP::OptNewtonLike::checkAnalyticFDGrad ( )

Compare the analytic gradient with the finite difference gradient.

virtual int OPTPP::OptNewtonLike::checkConvg ( ) [virtual]

Check to see if algorithm satisfies the convergence criterion.

Implements OPTPP::OptimizeClass.

virtual int OPTPP::OptNewtonLike::checkDeriv ( ) [virtual]

Compare the analytic gradient with the finite difference gradient.

Reimplemented in OPTPP::OptFDNewton, OPTPP::OptNewton, and OPTPP::OptQNewton.

virtual NEWMAT::ColumnVector OPTPP::OptNewtonLike::computeSearch ( NEWMAT::SymmetricMatrix & H ) [inline, virtual]

Compute Newton direction.

Implements OPTPP::OptimizeClass.

Definition at line 91 of file OptNewtonLike.h.

virtual int OPTPP::OptNewtonLike::computeStep ( NEWMAT::ColumnVector sk ) [virtual]

Compute the step length along the Newton direction.

void OPTPP::OptNewtonLike::defaultAcceptStep	(	int	,
		int
	)			`[protected, virtual]`

Provide default implementation of AcceptStep.

Reimplemented from OPTPP::OptimizeClass.

Referenced by acceptStep().

DerivOption OPTPP::OptNewtonLike::getDerivOption ( ) const [inline]

Set the type of finite difference routine.

Definition at line 186 of file OptNewtonLike.h.

References finitediff.

int OPTPP::OptNewtonLike::getFevals ( ) const [inline]

Returns:: The number of function evaluations

Definition at line 140 of file OptNewtonLike.h.

References OPTPP::OptimizeClass::fcn_evals.

int OPTPP::OptNewtonLike::getGevals ( ) const [inline]

Returns:: The number of gradient evaluations

Definition at line 145 of file OptNewtonLike.h.

References grad_evals.

real OPTPP::OptNewtonLike::getGradMult ( ) const [inline]

Returns:: Gradient multiplier to compute TR_size

Definition at line 157 of file OptNewtonLike.h.

References gradMult.

NEWMAT::SymmetricMatrix OPTPP::OptNewtonLike::getHessian ( ) const [inline]

Returns:: Hessian matrix

Definition at line 191 of file OptNewtonLike.h.

References Hessian.

int OPTPP::OptNewtonLike::getSearchSize ( ) const [inline]

Returns:: Number of points in search scheme which is used in trustpds search strategy

Definition at line 166 of file OptNewtonLike.h.

References searchSize.

SearchStrategy OPTPP::OptNewtonLike::getSearchStrategy ( ) const [inline]

Set globalization strategy for optimization algorithms.

Definition at line 178 of file OptNewtonLike.h.

References strategy.

real OPTPP::OptNewtonLike::getTRSize ( ) const [inline]

Returns:: Radius of trust-region

Definition at line 150 of file OptNewtonLike.h.

References TR_size.

virtual void OPTPP::OptNewtonLike::initHessian ( ) [virtual]

Compute the Hessian or its approximation at the initial point.

Reimplemented in OPTPP::OptNewton.

virtual void OPTPP::OptNewtonLike::initOpt ( ) [virtual]

Initialize algorithmic parameters.

virtual double OPTPP::OptNewtonLike::initTrustRegionSize ( ) const [virtual]

Initialize the size of the trust-region.

Only relevant when either the trustregion or trustpds globalization strategies are selected

virtual NLP1* OPTPP::OptNewtonLike::nlprob ( ) const [protected, pure virtual]

Returns pointer to an NLP1 object.

Implemented in OPTPP::OptNewton1Deriv, and OPTPP::OptNewton2Deriv.

virtual void OPTPP::OptNewtonLike::optimize ( ) [virtual]

Invoke Newton's method on an unconstrained problem.

Implements OPTPP::OptimizeClass.

void OPTPP::OptNewtonLike::printStatus ( char * ) [virtual]

Print status of unconstrained Newton's method.

Implements OPTPP::OptimizeClass.

Reimplemented in OPTPP::OptNewton.

virtual void OPTPP::OptNewtonLike::readOptInput ( ) [virtual]

Read user-specified input options from a file.

Implements OPTPP::OptimizeClass.

virtual void OPTPP::OptNewtonLike::reset ( ) [virtual]

Reset parameters.

Implements OPTPP::OptimizeClass.

void OPTPP::OptNewtonLike::setDerivOption ( DerivOption d ) [inline]

Returns:: Type of finite difference approximation.

Definition at line 184 of file OptNewtonLike.h.

References finitediff.

void OPTPP::OptNewtonLike::setGradMult ( real tau ) [inline]

Set gradient multiplier which is used to compute trust-region radius.

Definition at line 159 of file OptNewtonLike.h.

References gradMult.

void OPTPP::OptNewtonLike::setHessian ( NEWMAT::SymmetricMatrix & H ) [inline]

Store the current Hessian matrix.

Definition at line 193 of file OptNewtonLike.h.

References Hessian.

void OPTPP::OptNewtonLike::setSearchSize ( int sss ) [inline]

Set number of points in search scheme for trust-pds search strategy.

Definition at line 168 of file OptNewtonLike.h.

References searchSize.

void OPTPP::OptNewtonLike::setSearchStrategy ( SearchStrategy s ) [inline]

Returns:: Globalization strategy for optimization algorithms

Definition at line 176 of file OptNewtonLike.h.

References strategy.

void OPTPP::OptNewtonLike::setTRSize ( real delta ) [inline]

Set trust-region radius.

Definition at line 152 of file OptNewtonLike.h.

References TR_size.

virtual NEWMAT::SymmetricMatrix OPTPP::OptNewtonLike::updateH	(	NEWMAT::SymmetricMatrix &	H,
		int	k
	)			`[pure virtual]`

Compute the Hessian of the objective function or its approximation at the current point.

Implemented in OPTPP::OptFDNewton, OPTPP::OptNewton, and OPTPP::OptQNewton.

Member Data Documentation

DerivOption OPTPP::OptNewtonLike::finitediff [protected]

User-specified derivative option.

Definition at line 32 of file OptNewtonLike.h.

Referenced by getDerivOption(), and setDerivOption().

NEWMAT::ColumnVector OPTPP::OptNewtonLike::gprev [protected]

Gradient at the prev. iteration.

Definition at line 28 of file OptNewtonLike.h.

int OPTPP::OptNewtonLike::grad_evals [protected]

Number of gradient evaluations.

Definition at line 30 of file OptNewtonLike.h.

Referenced by getGevals().

real OPTPP::OptNewtonLike::gradMult [protected]

Gradient multiplier to compute TR_size.

Definition at line 34 of file OptNewtonLike.h.

Referenced by getGradMult(), and setGradMult().

NEWMAT::SymmetricMatrix OPTPP::OptNewtonLike::Hessian [protected]

Current Hessian.

Definition at line 29 of file OptNewtonLike.h.

Referenced by getHessian(), and setHessian().

int OPTPP::OptNewtonLike::searchSize [protected]

Search pattern size for TRPDS.

Definition at line 35 of file OptNewtonLike.h.

Referenced by getSearchSize(), and setSearchSize().

SearchStrategy OPTPP::OptNewtonLike::strategy [protected]

User-specified globalization Strategy.

Definition at line 31 of file OptNewtonLike.h.

Referenced by getSearchStrategy(), and setSearchStrategy().

real OPTPP::OptNewtonLike::TR_size [protected]

Trust region radius.

Definition at line 33 of file OptNewtonLike.h.

Referenced by getTRSize(), and setTRSize().

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

include/OptNewtonLike.h

OPTPP::OptNewtonLike Class Reference

Public Member Functions

Protected Member Functions

Protected Attributes

Friends

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation