MPQC  3.0.0-alpha
Public Member Functions | Protected Member Functions | Protected Attributes
sc::Function Class Reference

The Function class is an abstract base class that, given a set of coordinates, will compute a value and possibly a gradient and hessian at that point. More...

#include <math/optimize/function.h>

Inheritance diagram for sc::Function:
Inheritance graph
[legend]

List of all members.

Public Member Functions

 Function (StateIn &)
 Function (const Function &)
 Function (const Ref< KeyVal > &, double funcacc=DBL_EPSILON, double gradacc=DBL_EPSILON, double hessacc=DBL_EPSILON)
 The keyval constructor reads the following keywords:
Functionoperator= (const Function &)
Ref< SCMatrixKitmatrixkit () const
 Return the SCMatrixKit used to construct vectors and matrices.
RefSCDimension dimension () const
 Return the SCDimension of the problem.
virtual void save_data_state (StateOut &)
 Save the base classes (with save_data_state) and the members in the same order that the StateIn CTOR initializes them.
virtual double value ()
 Return the value of the function.
int value_needed () const
 Returns nonzero if the current value is not up-to-date.
int do_value (int)
 If passed a nonzero number, compute the value the next time compute() is called.
AccResultdoublevalue_result ()
virtual void set_desired_value_accuracy (double)
 Set the accuracy to which the value is to be computed.
virtual double actual_value_accuracy () const
 Return the accuracy with which the value has been computed.
virtual double desired_value_accuracy () const
 Return the accuracy with which the value is to be computed.
virtual void guess_hessian (RefSymmSCMatrix &)
 Compute a quick, approximate hessian.
virtual RefSymmSCMatrix inverse_hessian (RefSymmSCMatrix &)
virtual void set_x (const RefSCVector &)
 Set and retrieve the coordinate values.
RefSCVector get_x () const
const RefSCVectorget_x_no_copy () const
virtual Ref< NonlinearTransformchange_coordinates ()
 An optimizer can call change coordinates periodically to give the function an opportunity to change its coordinate system.
virtual void print (std::ostream &=ExEnv::out0()) const
 Print information about the object.
void print_desired_accuracy (std::ostream &=ExEnv::out0()) const
 similar to print(), but only prins desired accuracies
virtual bool throw_if_tolerance_exceeded () const
 Overridden Compute member.
Gradient Members

These are analogous to the routines that deal with values, but work with gradients instead.

virtual RefSCVector gradient ()
int gradient_needed () const
int do_gradient (int)
virtual void set_desired_gradient_accuracy (double)
virtual double actual_gradient_accuracy () const
virtual double desired_gradient_accuracy () const
AccResultRefSCVectorgradient_result ()
Hessian Members

These are analogous to the routines that deal with values, but work with the hessian instead.

virtual RefSymmSCMatrix hessian ()
int hessian_needed () const
int do_hessian (int)
virtual void set_desired_hessian_accuracy (double)
virtual double actual_hessian_accuracy () const
virtual double desired_hessian_accuracy () const
AccResultRefSymmSCMatrixhessian_result ()
Members that check whether the desired accuracies were set to the default values.
   
virtual bool desired_value_accuracy_set_to_default () const
virtual bool desired_gradient_accuracy_set_to_default () const
virtual bool desired_hessian_accuracy_set_to_default () const
Members that check the availability of values, gradients, and hessians.
     
virtual int value_implemented () const
virtual int gradient_implemented () const
virtual int hessian_implemented () const

Protected Member Functions

virtual void set_matrixkit (const Ref< SCMatrixKit > &)
 Set the SCMatrixKit that should be used to construct the requisite vectors and matrices.
virtual void set_dimension (const RefSCDimension &)
RefSCVectorget_x_reference ()
 Get read/write access to the coordinates for modification.
void do_change_coordinates (const Ref< NonlinearTransform > &)
 Change the coordinate system and apply the given transform to intermediates matrices and vectors.
Update Members

Update the various computable results.

virtual void set_value (double)
virtual void set_gradient (RefSCVector &)
virtual void set_hessian (RefSymmSCMatrix &)
Accuracy Setting Members

Set the accuracies with which the various computables have been computed.

virtual void set_actual_value_accuracy (double)
virtual void set_actual_gradient_accuracy (double)
virtual void set_actual_hessian_accuracy (double)

Protected Attributes

Ref< SCMatrixKitmatrixkit_
 Used to construct new matrices.
RefSCVector x_
 The variables.
RefSCDimension dim_
 The dimension of x_.
AccResultdouble value_
 The value of the function at x_.
AccResultRefSCVector gradient_
 The gradient at x_.
AccResultRefSymmSCMatrix hessian_
 The hessian at x_.
bool desired_value_accuracy_set_to_default_
bool desired_gradient_accuracy_set_to_default_
bool desired_hessian_accuracy_set_to_default_
bool throw_if_tolerance_exceeded_

Detailed Description

The Function class is an abstract base class that, given a set of coordinates, will compute a value and possibly a gradient and hessian at that point.


Constructor & Destructor Documentation

sc::Function::Function ( const Ref< KeyVal > &  ,
double  funcacc = DBL_EPSILON,
double  gradacc = DBL_EPSILON,
double  hessacc = DBL_EPSILON 
)

The keyval constructor reads the following keywords:

matrixkit Gives a SCMatrixKit object. If it is not specified, a default SCMatrixKit is selected.

value_accuracy Sets the accuracy to which values are computed. The default is the machine accuracy.

gradient_accuracy Sets the accuracy to which gradients are computed. The default is the machine accuracy.

hessian_accuracy Sets the accuracy to which hessians are computed. The default is the machine accuracy.

throw_if_tolerance_exceeded If this is true, then an exception will be thrown if a result cannot be computed to the desired accuracy. The default is true.


Member Function Documentation

An optimizer can call change coordinates periodically to give the function an opportunity to change its coordinate system.

A return value of 0 means the coordinates were not changed. Otherwise, a transform object to the new coordinate system is return. The function object applies the transform to any objects it contains. This will obsolete the function data.

Reimplemented in sc::MolecularEnergy.

void sc::Function::do_change_coordinates ( const Ref< NonlinearTransform > &  ) [protected]

Change the coordinate system and apply the given transform to intermediates matrices and vectors.

int sc::Function::do_value ( int  )

If passed a nonzero number, compute the value the next time compute() is called.

Return a nonzero number if the value was previously to be computed.

virtual int sc::Function::gradient_implemented ( ) const [virtual]
Returns:
1 (gradient implemented, analytically or numerically) or 0 (gradient not implemented, default).

Reimplemented in sc::UnionShape, sc::Uncapped5SphereExclusionShape, sc::MolecularEnergy, sc::ReentrantUncappedTorusHoleShape, sc::NonreentrantUncappedTorusHoleShape, sc::UncappedTorusHoleShape, and sc::SphereShape.

virtual int sc::Function::hessian_implemented ( ) const [virtual]
Returns:
1 (hessian implemented, analytically or numerically) or 0 (hessian not implemented, default).

Reimplemented in sc::MolecularEnergy.

Return the SCMatrixKit used to construct vectors and matrices.

virtual void sc::Function::save_data_state ( StateOut ) [virtual]
virtual void sc::Function::set_matrixkit ( const Ref< SCMatrixKit > &  ) [protected, virtual]

Set the SCMatrixKit that should be used to construct the requisite vectors and matrices.

virtual int sc::Function::value_implemented ( ) const [virtual]
Returns:
1 (value implemented) or 0 (value not implemented, default). Must be overridden for the class to be useful.

Reimplemented in sc::LMP2, sc::PT2R12, sc::MBPT2, sc::SumMolecularEnergy, sc::HCoreWfn, sc::MBPT2_R12, sc::CCR12, sc::HSOSKS, sc::SpinOrbitalPT2R12, sc::UKS, sc::MolecularEnergyCCA, sc::CLKS, sc::TaylorMolecularEnergy, sc::ExtendedHuckelWfn, sc::ExternPT2R12, sc::Shape, sc::UHF, sc::CLHF, sc::HSOSHF, sc::OSSHF, and sc::TCHF.


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

Generated at Sat Jul 7 2012 11:52:58 for MPQC 3.0.0-alpha using the documentation package Doxygen 1.8.0.