repicea.math.optimizer

Class NewtonRaphsonOptimizer

java.lang.Object

repicea.math.optimizer.AbstractOptimizer

repicea.math.optimizer.NewtonRaphsonOptimizer

public class NewtonRaphsonOptimizer
extends AbstractOptimizer

The NewtonRaphsonOptimizer class implements the Optimizer interface. It optimizes a log-likelihood function using the Newton-Raphson algorithm. The vector of parameter changes is estimated as - inv(d2LLK/d2Beta) * dLLK/dBeta.

Nested Class Summary

Nested classes/interfaces inherited from class repicea.math.optimizer.AbstractOptimizer

AbstractOptimizer.LineSearchMethod, AbstractOptimizer.OptimizationException

Field Summary

Fields

Modifier and Type	Field and Description
protected double	gradientCriterion
static java.lang.String	InnerIterationStarted
protected int	maxNumberOfIterations
protected boolean	verbose

Fields inherited from class repicea.math.optimizer.AbstractOptimizer

betaVector, convergenceAchieved, convergenceCriterion, hessianMatrix, optimalValue

Constructor Summary

Constructors

Constructor and Description
NewtonRaphsonOptimizer()

Method Summary

Modifier and Type	Method and Description
protected double	calculateConvergence(Matrix gradient, Matrix hessian, double llk) This method returns a double that is the convergence indicator based on the gradient, the hessian and the value of the log-likelihood function.
int	getNumberOfIterations() This method returns the number of iterations to reach convergence.
boolean	optimize(AbstractMathematicalFunction function, java.util.List<java.lang.Integer> indicesOfParametersToOptimize) This method optimizes the likelihood of the StatisticalModel instance.
protected double	runInnerOptimisation(AbstractMathematicalFunction function, java.util.List<java.lang.Integer> indicesOfParametersToOptimize, Matrix originalBeta, Matrix optimisationStep, double previousLogLikelihood, AbstractOptimizer.LineSearchMethod lineSearchMethod) This method optimize the log-likelihood function using the Newton-Raphson optimisation step.
void	setVerbose(boolean verbose) This method enables the verbose, i.e.

Methods inherited from class repicea.math.optimizer.AbstractOptimizer

addOptimizerListener, fireOptimizerEvent, getConvergenceCriterion, getHessianAtMaximum, getOptimalValue, getParametersAtMaximum, isConvergenceAchieved, isVerboseEnabled, removeOptimizerListener, setConvergenceCriterion, setVerboseEnabled, toString

Methods inherited from class java.lang.Object

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

Field Detail

InnerIterationStarted

public static final java.lang.String InnerIterationStarted

See Also:

Constant Field Values

verbose

protected boolean verbose

maxNumberOfIterations

protected int maxNumberOfIterations

gradientCriterion

protected double gradientCriterion

Constructor Detail

NewtonRaphsonOptimizer

public NewtonRaphsonOptimizer()

Method Detail

runInnerOptimisation

protected double runInnerOptimisation(AbstractMathematicalFunction function,
                                      java.util.List<java.lang.Integer> indicesOfParametersToOptimize,
                                      Matrix originalBeta,
                                      Matrix optimisationStep,
                                      double previousLogLikelihood,
                                      AbstractOptimizer.LineSearchMethod lineSearchMethod)
                               throws AbstractOptimizer.OptimizationException

This method optimize the log-likelihood function using the Newton-Raphson optimisation step.

Parameters:

function - an AbstractMathematicalFunction instance

indicesOfParametersToOptimize - a list of the indices of the parameters to be optimized

originalBeta - the vector that contains the parameters of the previous outer optimisation

optimisationStep - the optimisation step from the Newton-Raphson algorithm

previousLogLikelihood - the value of the log-likelihood function in the last outer optimisation

lineSearchMethod -

Returns:

the value of the function

Throws:

OptimisationException - if the inner optimisation is not successful

AbstractOptimizer.OptimizationException

optimize

public boolean optimize(AbstractMathematicalFunction function,
                        java.util.List<java.lang.Integer> indicesOfParametersToOptimize)
                 throws AbstractOptimizer.OptimizationException

Description copied from class: AbstractOptimizer

This method optimizes the likelihood of the StatisticalModel instance.

Specified by:

optimize in class AbstractOptimizer

Parameters:

function - a AbstractMathematicalFunction object

indicesOfParametersToOptimize - a List instance that contains the indices of the parameters to be optimized

Throws:

AbstractOptimizer.OptimizationException

calculateConvergence

protected double calculateConvergence(Matrix gradient,
                                      Matrix hessian,
                                      double llk)

This method returns a double that is the convergence indicator based on the gradient, the hessian and the value of the log-likelihood function.

Parameters:

gradient - a Matrix instance

hessian - a Matrix instance

llk - a double that is the value of the log-likelihood function

Returns:

the indicator (a Double instance)

getNumberOfIterations

public int getNumberOfIterations()

This method returns the number of iterations to reach convergence. It returns -1 if convergence could not be reached.

Returns:

an integer

setVerbose

public void setVerbose(boolean verbose)

This method enables the verbose, i.e. some messages will be displayed in the console. The default verbose value is false.

Parameters:

verbose - a boolean