OPTPP::OptConstrNewtonLike Class Reference

#include <OptConstrNewtonLike.h>

Inheritance diagram for OPTPP::OptConstrNewtonLike:

List of all members.

Public Member Functions
	OptConstrNewtonLike ()
	OptConstrNewtonLike (int n)
	OptConstrNewtonLike (int n, UPDATEFCN u)
	OptConstrNewtonLike (int n, TOLS t)
virtual	~OptConstrNewtonLike ()
virtual void	acceptStep (int k, int step_type)
	Accept this step and update the nonlinear model.
virtual NEWMAT::ColumnVector	computeSearch (NEWMAT::SymmetricMatrix &H)
	Solve for the Newton direction.
virtual void	updateModel (int k, int ndim, NEWMAT::ColumnVector x)
virtual void	reset ()
virtual int	checkConvg ()
	Check to see if algorithm satisfies the convergence criterion.
virtual int	checkDeriv ()
	Compare the analytic gradient with the finite-difference approximation.
virtual int	computeStep (NEWMAT::ColumnVector sk)
	Compute the steplength along the search direction. If an acceptable step not found, returns an error code = -1.
virtual real	computeMaxStep (NEWMAT::ColumnVector sk)
virtual void	initOpt ()
	Initialize algorithmic parameters.
virtual void	initHessian ()
	Compute the Hessian or its approximation at the initial point.
virtual double	initTrustRegionSize () const
	Initialize the radius of the trust-region.
virtual void	optimize ()
	Invoke a constrained Newton's method.
virtual void	readOptInput ()
	Read user-specified input options from a file.
int	checkAnalyticFDGrad ()
	Check finite difference approximation with analytic derivatives.
int	getFevals () const
int	getGevals () const
real	getTRSize () const
void	setTRSize (real delta)
	Set radius of trust-region.
real	getGradMult () const
void	setGradMult (real tau)
	Set gradient multiplier which is used to compute radius of trust-region.
int	getSearchSize () const
void	setSearchSize (int sss)
	Set number of points in search scheme for trustpds strategy.
bool	getWarmStart () const
void	UseWarmStart (NEWMAT::SymmetricMatrix &H)
void	setSearchStrategy (SearchStrategy search)
SearchStrategy	getSearchStrategy () const
	Set globalization strategy for optimization algorithm.
void	setDerivOption (DerivOption d)
DerivOption	getDerivOption () const
	Set type of finite difference approximation (forward, backward, or central).
NEWMAT::ColumnVector	getEqualityMultiplier () const
void	setEqualityMultiplier (const NEWMAT::ColumnVector &ymult)
	Store the current Lagrangian multipliers corresponding to equality constraints.
NEWMAT::ColumnVector	getInequalityMultiplier () const
void	setInequalityMultiplier (const NEWMAT::ColumnVector &zmult)
	Store the current Lagrangian multipliers corresponding to inequality constraints.
NEWMAT::ColumnVector	getSlacks () const
void	setSlacks (const NEWMAT::ColumnVector &slackVar)
	Store the current slack vector.
MeritFcn	getMeritFcn () const
virtual void	setMeritFcn (MeritFcn option)
	Specify the merit function to used in step acceptance test.
NEWMAT::ColumnVector	getGradL () const
virtual void	setGradL (NEWMAT::ColumnVector gradl_value)
	Store the gradient of the Lagrangian at the current iteration.
NEWMAT::ColumnVector	getGradLPrev () const
virtual void	setGradLPrev (NEWMAT::ColumnVector gradl_value)
	Store the gradient of the Lagrangian at the previous iteration.
NEWMAT::ColumnVector	getConstraintResidual () const
virtual void	setConstraintResidual (const NEWMAT::ColumnVector &constraint_value)
	Store the residuals of the constraints at the current iteration.
NEWMAT::Matrix	getConstraintGradient () const
virtual void	setConstraintGradient (const NEWMAT::Matrix &constraint_grad)
	Store the current gradients of the constraints.
real	getCost () const
void	setCost (real value)
	Store current value of merit function.
bool	getFeasibilityRecoveryFlag () const
void	setFeasibilityRecoveryFlag (bool flag)
	Set switch to turn on feasibility recovery method.
int	getMaxFeasIter () const
void	setMaxFeasIter (int k)
	Set maximum number of iterations for feasibility recovery method of trust-region.
NEWMAT::SymmetricMatrix	getHessian () const
void	setHessian (NEWMAT::SymmetricMatrix &H)
	Store current Hessian matrix.
virtual NEWMAT::SymmetricMatrix	updateH (NEWMAT::SymmetricMatrix &H, int k)=0
	Compute the Hessian of the Lagrangrian or its approximation at iteration k.
NEWMAT::ColumnVector	computeFFK1Ind (const NEWMAT::ColumnVector &xc)
	Compute the active set according to Facchinei, Fischer, and Kanzow indicator.
NEWMAT::ColumnVector	computeFFK2Ind (const NEWMAT::ColumnVector &xc)
	Compute the active set according to Facchinei, Fischer, and Kanzow indicator.
NEWMAT::ColumnVector	computeTapiaInd (const NEWMAT::ColumnVector &step)
	Compute the Tapia indicators for the constrained problem.
void	printStatus (char *)
	Print status of the constrained Newton's method.
void	printMultipliers (char *)
	Output the Lagrangian multipliers to the screen.
void	fPrintMultipliers (ostream *nlpout, char *)
	Print the Lagrangian multipliers to a file.
void	fPrintSecSuff (ostream *nlpout, NEWMAT::ColumnVector &info)
	Print second order sufficiency information to a file.
Protected Member Functions
virtual NLP1 *	nlprob () const =0
	returns an NLP1 pointer
void	defaultAcceptStep (int, int)
NEWMAT::ColumnVector	defaultComputeSearch (NEWMAT::SymmetricMatrix &)
Protected Attributes
int	me
	Number of equality constraints.
int	mi
	Number of inequality constraints.
int	grad_evals
	Number of gradient evaluations.
NEWMAT::ColumnVector	gprev
	Gradient at previous iteration.
NEWMAT::ColumnVector	z
	Lagrange mult. corr. to ineq constraints.
NEWMAT::ColumnVector	y
	Lagrange mult. corr. to eq constraints.
NEWMAT::ColumnVector	s
	Slack variables corr. to ineq. constraints.
NEWMAT::ColumnVector	constrType
	Vector which contains the type of constraint.
NEWMAT::ColumnVector	constraintResidual
	Constraint residual at xc.
NEWMAT::ColumnVector	gradl
	Current gradient of the lagrangian.
NEWMAT::ColumnVector	gradlprev
	Previous gradient of the lagrangian.
NEWMAT::Matrix	constraintGradient
	Current constraint gradient.
NEWMAT::Matrix	constraintGradientPrev
	Previous constraint gradient.
NEWMAT::SymmetricMatrix	Hessian
	Current Hessian.
NEWMAT::SymmetricMatrix	hessl
	Current Hessian of the lagrangian.
SearchStrategy	strategy
	User-specified globalization strategy.
DerivOption	finitediff
	User-specified derivative option.
MeritFcn	mfcn
	Merit function option.
real	TR_size
	Size of the trust region radius.
real	gradMult
	Gradient multiplier to compute TR_size.
int	searchSize
	Search pattern size for TRPDS.
real	cost
	Value of the merit function.
bool	WarmStart
bool	feas_flag
	Switch to turn on the feasibility recovery method.
int	max_feas_iter
	Maximize number of iterations to achieve feasibility.
Friends
int	trustregion (NLP1 *, ostream *, NEWMAT::SymmetricMatrix &, NEWMAT::ColumnVector &, NEWMAT::ColumnVector &, real &, real &, real stpmax, real stpmin)
int	trustpds (NLP1 *, ostream *, NEWMAT::SymmetricMatrix &, NEWMAT::ColumnVector &, NEWMAT::ColumnVector &, real &, real &, real stpmax, real stpmin, int)

---

Detailed Description

Constrained Newton abstract data classes OptConstrNewtonLike OptConstrNewton1Deriv OptConstrNewton2Deriv

OptConstrNewtonLike provides common data and functionality for OptConstrFDNewton, OptConstrQNewton, OptConstrNewton, OptFDNIPS, OptQNIPS, and OptNIPS methods.

Author:

J.C. Meza, Lawrence Berkeley National Laboratory

Note:

Modified by P.J. Williams

Date:

03/27/2007

---

Constructor & Destructor Documentation

OPTPP::OptConstrNewtonLike::OptConstrNewtonLike (   )  [inline]

Default Constructor See also: OptConstrNewtonLike(int n) OptConstrNewtonLike(int n, UPDATEFCN u) OptConstrNewtonLike(int n, TOLS t)

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

Parameters: n  an integer argument See also: OptConstrNewtonLike(int n, UPDATEFCN u) OptConstrNewtonLike(int n, TOLS t)

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

Parameters: n  an integer argument u  a function pointer. See also: OptConstrNewtonLike(int n) OptConstrNewtonLike(int n, TOLS t)

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

Parameters: n  an integer argument t  tolerance class reference. See also: OptConstrNewtonLike(int n) OptConstrNewtonLike(int n, UPDATEFCN u)

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

Destructor

---

Member Function Documentation

virtual void OPTPP::OptConstrNewtonLike::acceptStep (  int  k, int  step_type )  [inline, virtual]

Accept this step and update the nonlinear model. Implements OPTPP::OptimizeClass.

int OPTPP::OptConstrNewtonLike::checkAnalyticFDGrad (   )

Check finite difference approximation with analytic derivatives.

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

Check to see if algorithm satisfies the convergence criterion. Implements OPTPP::OptimizeClass. Reimplemented in OPTPP::OptNIPSLike.

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

Compare the analytic gradient with the finite-difference approximation. Reimplemented in OPTPP::OptConstrFDNewton, OPTPP::OptConstrNewton, OPTPP::OptConstrQNewton, OPTPP::OptFDNIPS, OPTPP::OptNIPSLike, and OPTPP::OptQNIPS.

ColumnVector OPTPP::OptConstrNewtonLike::computeFFK1Ind (  const NEWMAT::ColumnVector &  xc  )

Compute the active set according to Facchinei, Fischer, and Kanzow indicator.

ColumnVector OPTPP::OptConstrNewtonLike::computeFFK2Ind (  const NEWMAT::ColumnVector &  xc  )

Compute the active set according to Facchinei, Fischer, and Kanzow indicator.

real OPTPP::OptConstrNewtonLike::computeMaxStep (  NEWMAT::ColumnVector  sk  )  [virtual]

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

Solve for the Newton direction. Implements OPTPP::OptimizeClass.

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

Compute the steplength along the search direction. If an acceptable step not found, returns an error code = -1. Reimplemented in OPTPP::OptNIPSLike.

ColumnVector OPTPP::OptConstrNewtonLike::computeTapiaInd (  const NEWMAT::ColumnVector &  step  )

Compute the Tapia indicators for the constrained problem.

void OPTPP::OptConstrNewtonLike::defaultAcceptStep (  int  iter, int  step_type )  [protected, virtual]

Provide default implementation of AcceptStep Reimplemented from OPTPP::OptimizeClass.

ColumnVector OPTPP::OptConstrNewtonLike::defaultComputeSearch (  NEWMAT::SymmetricMatrix &   )  [protected]

void OPTPP::OptConstrNewtonLike::fPrintMultipliers (  ostream *  nlpout, char *  )

Print the Lagrangian multipliers to a file.

void OPTPP::OptConstrNewtonLike::fPrintSecSuff (  ostream *  nlpout, NEWMAT::ColumnVector &  info )

Print second order sufficiency information to a file.

NEWMAT::Matrix OPTPP::OptConstrNewtonLike::getConstraintGradient (   )  const [inline]

Returns: Gradient of the constraints at the current iteration

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::getConstraintResidual (   )  const [inline]

Returns: Residuals of the constraints at the current iteration

real OPTPP::OptConstrNewtonLike::getCost (   )  const [inline]

Returns: Value of merit function

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

Set type of finite difference approximation (forward, backward, or central).

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::getEqualityMultiplier (   )  const [inline]

Returns: Lagrangian multipliers corresponding to equality constraints

bool OPTPP::OptConstrNewtonLike::getFeasibilityRecoveryFlag (   )  const [inline]

Returns: Switch to turn on feasibility recovery method

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

Returns: The number of function evaluations

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

Returns: The number of gradient evaluations

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::getGradL (   )  const [inline]

Returns: Gradient of the Lagrangian at the current iteration

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::getGradLPrev (   )  const [inline]

Returns: Gradient of the Lagrangian at the previous iteration

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

Returns: Gradient multiplier to compute radius of trust-region

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

Returns: Hessian matrix

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::getInequalityMultiplier (   )  const [inline]

Returns: Lagrangian multipliers corresponding to inequality constraints

int OPTPP::OptConstrNewtonLike::getMaxFeasIter (   )  const [inline]

Returns: Maximum number of iterations for feasibility recovery method

MeritFcn OPTPP::OptConstrNewtonLike::getMeritFcn (   )  const [inline]

Returns: Merit function

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

Returns: Number of points in search scheme (only relevant when trustpds search strategy is selected)

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

Set globalization strategy for optimization algorithm.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::getSlacks (   )  const [inline]

Returns: Slack variables associated with inequality constraints

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

Returns: Trust region radius

bool OPTPP::OptConstrNewtonLike::getWarmStart (   )  const [inline]

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

Compute the Hessian or its approximation at the initial point. Reimplemented in OPTPP::OptConstrNewton, OPTPP::OptDHNIPS, OPTPP::OptNIPS, and OPTPP::OptNIPSLike.

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

Initialize algorithmic parameters. Reimplemented in OPTPP::OptNIPSLike.

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

Initialize the radius of the trust-region.

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

returns an NLP1 pointer Implemented in OPTPP::OptConstrNewton1Deriv, OPTPP::OptConstrNewton2Deriv, OPTPP::OptDHNIPS, OPTPP::OptFDNIPS, OPTPP::OptNIPS, OPTPP::OptNIPSLike, and OPTPP::OptQNIPS.

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

Invoke a constrained Newton's method. Implements OPTPP::OptimizeClass. Reimplemented in OPTPP::OptNIPSLike.

void OPTPP::OptConstrNewtonLike::printMultipliers (  char *   )

Output the Lagrangian multipliers to the screen.

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

Print status of the constrained Newton's method. Implements OPTPP::OptimizeClass. Reimplemented in OPTPP::OptConstrNewton, OPTPP::OptDHNIPS, OPTPP::OptNIPS, and OPTPP::OptNIPSLike.

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

Read user-specified input options from a file. Implements OPTPP::OptimizeClass. Reimplemented in OPTPP::OptNIPSLike.

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

Implements OPTPP::OptimizeClass. Reimplemented in OPTPP::OptDHNIPS, and OPTPP::OptNIPSLike.

virtual void OPTPP::OptConstrNewtonLike::setConstraintGradient (  const NEWMAT::Matrix &  constraint_grad  )  [inline, virtual]

Store the current gradients of the constraints.

virtual void OPTPP::OptConstrNewtonLike::setConstraintResidual (  const NEWMAT::ColumnVector &  constraint_value  )  [inline, virtual]

Store the residuals of the constraints at the current iteration.

void OPTPP::OptConstrNewtonLike::setCost (  real  value  )  [inline]

Store current value of merit function.

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

Returns: Type of finite difference approximation

void OPTPP::OptConstrNewtonLike::setEqualityMultiplier (  const NEWMAT::ColumnVector &  ymult  )  [inline]

Store the current Lagrangian multipliers corresponding to equality constraints.

void OPTPP::OptConstrNewtonLike::setFeasibilityRecoveryFlag (  bool  flag  )  [inline]

Set switch to turn on feasibility recovery method.

virtual void OPTPP::OptConstrNewtonLike::setGradL (  NEWMAT::ColumnVector  gradl_value  )  [inline, virtual]

Store the gradient of the Lagrangian at the current iteration.

virtual void OPTPP::OptConstrNewtonLike::setGradLPrev (  NEWMAT::ColumnVector  gradl_value  )  [inline, virtual]

Store the gradient of the Lagrangian at the previous iteration.

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

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

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

Store current Hessian matrix.

void OPTPP::OptConstrNewtonLike::setInequalityMultiplier (  const NEWMAT::ColumnVector &  zmult  )  [inline]

Store the current Lagrangian multipliers corresponding to inequality constraints.

void OPTPP::OptConstrNewtonLike::setMaxFeasIter (  int  k  )  [inline]

Set maximum number of iterations for feasibility recovery method of trust-region.

virtual void OPTPP::OptConstrNewtonLike::setMeritFcn (  MeritFcn  option  )  [inline, virtual]

Specify the merit function to used in step acceptance test. Reimplemented in OPTPP::OptNIPSLike.

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

Set number of points in search scheme for trustpds strategy.

void OPTPP::OptConstrNewtonLike::setSearchStrategy (  SearchStrategy  search  )  [inline]

Returns: Globalization strategy for optimization algorithm

void OPTPP::OptConstrNewtonLike::setSlacks (  const NEWMAT::ColumnVector &  slackVar  )  [inline]

Store the current slack vector.

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

Set radius of trust-region.

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

Compute the Hessian of the Lagrangrian or its approximation at iteration k. Implemented in OPTPP::OptConstrFDNewton, OPTPP::OptConstrNewton, OPTPP::OptConstrQNewton, OPTPP::OptDHNIPS, OPTPP::OptFDNIPS, OPTPP::OptNIPS, OPTPP::OptNIPSLike, and OPTPP::OptQNIPS.

virtual void OPTPP::OptConstrNewtonLike::updateModel (  int  k, int  ndim, NEWMAT::ColumnVector  x )  [inline, virtual]

Implements OPTPP::OptimizeClass.

void OPTPP::OptConstrNewtonLike::UseWarmStart (  NEWMAT::SymmetricMatrix &  H  )  [inline]

---

Friends And Related Function Documentation

int trustpds (  NLP1 *  , ostream *  , NEWMAT::SymmetricMatrix &  , NEWMAT::ColumnVector &  , NEWMAT::ColumnVector &  , real &  , real &  , real  stpmax, real  stpmin, int  )  [friend]

int trustregion (  NLP1 *  , ostream *  , NEWMAT::SymmetricMatrix &  , NEWMAT::ColumnVector &  , NEWMAT::ColumnVector &  , real &  , real &  , real  stpmax, real  stpmin )  [friend]

---

Member Data Documentation

NEWMAT::Matrix OPTPP::OptConstrNewtonLike::constraintGradient [protected]

Current constraint gradient.

NEWMAT::Matrix OPTPP::OptConstrNewtonLike::constraintGradientPrev [protected]

Previous constraint gradient.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::constraintResidual [protected]

Constraint residual at xc.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::constrType [protected]

Vector which contains the type of constraint.

real OPTPP::OptConstrNewtonLike::cost [protected]

Value of the merit function.

bool OPTPP::OptConstrNewtonLike::feas_flag [protected]

Switch to turn on the feasibility recovery method.

DerivOption OPTPP::OptConstrNewtonLike::finitediff [protected]

User-specified derivative option.

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

Gradient at previous iteration.

int OPTPP::OptConstrNewtonLike::grad_evals [protected]

Number of gradient evaluations.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::gradl [protected]

Current gradient of the lagrangian.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::gradlprev [protected]

Previous gradient of the lagrangian.

real OPTPP::OptConstrNewtonLike::gradMult [protected]

Gradient multiplier to compute TR_size.

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

Current Hessian.

NEWMAT::SymmetricMatrix OPTPP::OptConstrNewtonLike::hessl [protected]

Current Hessian of the lagrangian.

int OPTPP::OptConstrNewtonLike::max_feas_iter [protected]

Maximize number of iterations to achieve feasibility.

int OPTPP::OptConstrNewtonLike::me [protected]

Number of equality constraints.

MeritFcn OPTPP::OptConstrNewtonLike::mfcn [protected]

Merit function option.

int OPTPP::OptConstrNewtonLike::mi [protected]

Number of inequality constraints.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::s [protected]

Slack variables corr. to ineq. constraints.

int OPTPP::OptConstrNewtonLike::searchSize [protected]

Search pattern size for TRPDS.

SearchStrategy OPTPP::OptConstrNewtonLike::strategy [protected]

User-specified globalization strategy.

real OPTPP::OptConstrNewtonLike::TR_size [protected]

Size of the trust region radius.

bool OPTPP::OptConstrNewtonLike::WarmStart [protected]

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::y [protected]

Lagrange mult. corr. to eq constraints.

NEWMAT::ColumnVector OPTPP::OptConstrNewtonLike::z [protected]

Lagrange mult. corr. to ineq constraints.

---

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

• OptConstrNewtonLike.h
• OptConstrNewtonLike.C

---

Bug Reports    OPT++ Developers    Copyright Information    GNU Lesser General Public License Documentation, generated by , last revised August 30, 2006.