org.omegahat.Numerics.Optimizers
Class OptimizerAlgorithmBasic

java.lang.Object
  |
  +--org.omegahat.Numerics.Optimizers.OptimizerAlgorithmBasic

All Implemented Interfaces:

OptimizerAlgorithm

Direct Known Subclasses:

OptimizerAlgorithmBFGS

public abstract class OptimizerAlgorithmBasic

extends java.lang.Object

implements OptimizerAlgorithm

A class with some basic methods for any optimizer.

Mostly, this provides utilities, such as for handling the state, which can be used by algorithms with specific numerical methods.

Field Summary

protected java.lang.Class	arrayClass
protected NumericArray	arrayTemplate An object of the class to be used for NumericMatrix objects in the optimizer algorithm.
protected boolean	gradient Does this algorithm expect gradients to be computed for model points?
protected boolean	hessian Does this algorithm expect hessians (the matrix of second partial derivatives) to be computed for model points?
protected java.lang.Class	matrixClass
protected DynamicPackage	matrixPackage The package from which classes implementing NumericMatrix and NumericArray will be extracted.
protected NumericMatrix	matrixTemplate An object of the class to be used for NumericMatrix objects in the optimizer algorithm.
protected ModelObjective	model
protected java.lang.Class	modelPointClass
protected ModelPoint	modelPointTemplate
protected ModelPointNumericInt	start The starting point for the optimization.
protected int	state

Constructor Summary

OptimizerAlgorithmBasic()

Method Summary

java.lang.Class	getArrayClass() Accessor for arrayClass field
NumericArray	getArrayTemplate() Accessor for arrayTemplate field
java.lang.Class	getMatrixClass() Accessor for matrixClass field
DynamicPackage	getMatrixPackage() Accessor for matrixPackage field
NumericMatrix	getMatrixTemplate() Accessor for matrixTemplate field
ModelObjective	getModel() Accessor for model field
java.lang.Class	getModelPointClass() Accessor for modelPointClass field
ModelPoint	getModelPointTemplate() Accessor for modelPointTemplate field
ModelPointNumericInt	getStart() Accessor for start field
int	getState() Accessor for state field
boolean	hasGradient() Accessor for gradient field
boolean	hasHessian() Accessor for hessian field
ModelPoint	initialize(double[] parameters, OptimizerIterator opt) Initialize the iteration with these numerical values for the parameter estimates.
ModelPoint	initialize(ModelPoint theta, OptimizerIterator i) Initialize from a ModelPoint starting value and an iterator.
NumericArray	newArray() Generate a new object from the NumericArray class defined by the matrixPackage for this object.
NumericMatrix	newMatrix() Generate a new object from the NumericMatrix class defined by the matrixPackage for this object.
ModelPoint	newModelPoint() Generate a new object from the ModelPoint class defined by the matrixPackage for this object.
java.lang.Class	setArrayClass(java.lang.Class value) Accessor for setting arrayClass field
NumericArray	setArrayTemplate(NumericArray value) Accessor for setting arrayTemplate field
boolean	setGradient(boolean value) Accessor for setting gradient field
boolean	setHessian(boolean value) Accessor for setting hessian field
java.lang.Class	setMatrixClass(java.lang.Class value) Accessor for setting matrixClass field
DynamicPackage	setMatrixPackage() Establish the actual classes corresponding to the current matrixPackage property.
DynamicPackage	setMatrixPackage(DynamicPackage value) Accessor for setting matrixPackage field from a LinearAlgebra.Package object.
DynamicPackage	setMatrixPackage(java.lang.String packageName) Set the package to that corresponding to the specified name.
NumericMatrix	setMatrixTemplate(NumericMatrix value) Accessor for setting matrixTemplate field
ModelObjective	setModel(ModelObjective value) Accessor for setting model field
java.lang.Class	setModelPointClass(java.lang.Class value) Accessor for setting modelPointClass field
ModelPoint	setModelPointTemplate(ModelPoint value) Accessor for setting modelPointTemplate field
ModelPointNumericInt	setStart(ModelPointNumericInt value) Accessor for setting start field.
ModelPointNumericInt	setStart(NumericArray value) Set the start from a numeric array of parameter values.
int	setState(int value) Accessor for setting state field
ModelPoint	wrapup(ModelPoint pars, OptimizerIterator opt) Wrap up the completed iteration.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface org.omegahat.Numerics.Optimizers.OptimizerAlgorithm

refine

Field Detail

model

protected ModelObjective model

start

protected ModelPointNumericInt start

The starting point for the optimization.

This property is normally set during initialization. If the specified ModelPoint includes the definition of the objective function, setting the start property defines the model objective as well.

state

protected int state

matrixPackage

protected DynamicPackage matrixPackage

The package from which classes implementing NumericMatrix and NumericArray will be extracted.

gradient

protected boolean gradient

Does this algorithm expect gradients to be computed for model points?

hessian

protected boolean hessian

Does this algorithm expect hessians (the matrix of second partial derivatives) to be computed for model points?

matrixClass

protected java.lang.Class matrixClass

arrayClass

protected java.lang.Class arrayClass

modelPointClass

protected java.lang.Class modelPointClass

modelPointTemplate

protected ModelPoint modelPointTemplate

matrixTemplate

protected NumericMatrix matrixTemplate

An object of the class to be used for NumericMatrix objects in the optimizer algorithm. Usually set up by the first initialize call to the algoirthm. See matrixPackage for the ways the package or class can be specified.

arrayTemplate

protected NumericArray arrayTemplate

An object of the class to be used for NumericMatrix objects in the optimizer algorithm. Usually set up by the first initialize call to the algoirthm. See matrixPackage for the ways the package or class can be specified.

Constructor Detail

OptimizerAlgorithmBasic

public OptimizerAlgorithmBasic()

Method Detail

getModel

public ModelObjective getModel()

Accessor for model field

setModel

public ModelObjective setModel(ModelObjective value)

Accessor for setting model field

getStart

public ModelPointNumericInt getStart()

Accessor for start field

setStart

public ModelPointNumericInt setStart(ModelPointNumericInt value)

Accessor for setting start field.

setStart

public ModelPointNumericInt setStart(NumericArray value)

Set the start from a numeric array of parameter values.

Uses the matrix package specified for this optimizer algorithm to create the internal start object with parameters of the same class the algorithm will use, independent of the arguments to setStart.

getState

public int getState()

Accessor for state field

setState

public int setState(int value)

Accessor for setting state field

getMatrixPackage

public DynamicPackage getMatrixPackage()

Accessor for matrixPackage field

setMatrixPackage

public DynamicPackage setMatrixPackage(DynamicPackage value)

Accessor for setting matrixPackage field from a LinearAlgebra.Package object.

hasGradient

public boolean hasGradient()

Accessor for gradient field

setGradient

public boolean setGradient(boolean value)

Accessor for setting gradient field

hasHessian

public boolean hasHessian()

Accessor for hessian field

setHessian

public boolean setHessian(boolean value)

Accessor for setting hessian field

getMatrixClass

public java.lang.Class getMatrixClass()

Accessor for matrixClass field

setMatrixClass

public java.lang.Class setMatrixClass(java.lang.Class value)

Accessor for setting matrixClass field

getArrayClass

public java.lang.Class getArrayClass()

Accessor for arrayClass field

setArrayClass

public java.lang.Class setArrayClass(java.lang.Class value)

Accessor for setting arrayClass field

getModelPointClass

public java.lang.Class getModelPointClass()

Accessor for modelPointClass field

setModelPointClass

public java.lang.Class setModelPointClass(java.lang.Class value)

Accessor for setting modelPointClass field

getModelPointTemplate

public ModelPoint getModelPointTemplate()

Accessor for modelPointTemplate field

setModelPointTemplate

public ModelPoint setModelPointTemplate(ModelPoint value)

Accessor for setting modelPointTemplate field

getMatrixTemplate

public NumericMatrix getMatrixTemplate()

Accessor for matrixTemplate field

setMatrixTemplate

public NumericMatrix setMatrixTemplate(NumericMatrix value)

Accessor for setting matrixTemplate field

getArrayTemplate

public NumericArray getArrayTemplate()

Accessor for arrayTemplate field

setArrayTemplate

public NumericArray setArrayTemplate(NumericArray value)

Accessor for setting arrayTemplate field

initialize

public ModelPoint initialize(double[] parameters,
                             OptimizerIterator opt)

Initialize the iteration with these numerical values for the parameter estimates.

The model to be used for the start property must have been defined before this initialize method is called, since only the numeric values of the parameters are supplied here.

initialize

public ModelPoint initialize(ModelPoint theta,
                             OptimizerIterator i)

Initialize from a ModelPoint starting value and an iterator. The model point object defines both the form of the model and the numerical values to use as starting estimates.

The returned value is not necessarily identical to the argument, because the particular NumericArray class used to store the parameter values must match that needed for the computations of this algorithm. The initializer will in this case replace the parameter estimates in theta with an array of the correct class, but with the same numerical values.

Specified by:

initialize in interface OptimizerAlgorithm

wrapup

public ModelPoint wrapup(ModelPoint pars,
                         OptimizerIterator opt)

Wrap up the completed iteration.

The basic method does nothing. Over-riding methods may want to complete some estimates at the convergent parameter values (after testing the state of the optimization iterator against OptimizerIterator.CONVERGED).

Specified by:

wrapup in interface OptimizerAlgorithm

setMatrixPackage

public DynamicPackage setMatrixPackage(java.lang.String packageName)

Set the package to that corresponding to the specified name.

Individual classes then have names corresponding to the convention of appending the package name to the corresponding interface's name (see DynamicPackage documentation for details).

setMatrixPackage

public DynamicPackage setMatrixPackage()

Establish the actual classes corresponding to the current matrixPackage property. See matrixPackage for the options.

newMatrix

public NumericMatrix newMatrix()

Generate a new object from the NumericMatrix class defined by the matrixPackage for this object.

newArray

public NumericArray newArray()

Generate a new object from the NumericArray class defined by the matrixPackage for this object.

newModelPoint

public ModelPoint newModelPoint()

Generate a new object from the ModelPoint class defined by the matrixPackage for this object.

Specified by:

newModelPoint in interface OptimizerAlgorithm