meshi.optimizers
Class ConjugateGradient

java.lang.Object
  meshi.optimizers.Optimizer
      meshi.optimizers.Minimizer
          meshi.optimizers.ConjugateGradient

public class ConjugateGradient
extends Minimizer

This class implements a nonlinear Conjugate Gradient minimizer PR+ algorithm is implemented. Fletcher-Reeves algorithm is the base (FR-CG 5.4 in book below) using betaPR (PR-CG algorithm) (Polak-Ribiere) 5.43 and using 5.44 (non-negative beta: PR+) pp. 122. with posible restarts every n steps. according to the scheme in: Numerical Optimization by J. Nocendal & S. J. Wright, Springer 1999, pp 120-122.

Nested Class Summary

Nested classes/interfaces inherited from class meshi.optimizers.Optimizer

Optimizer.OptimizerStatus

Field Summary

protected double	beta
protected double[][]	bufferCoordinates
private double	c1
private double	c2
protected double[][]	coordinates
private static double	DEFAULT_C1
private static double	DEFAULT_C2
private static double	DEFAULT_EXTENDED_ALPHA_FACTOR_WOLF_SEARCH
private static int	DEFAULT_MAX_ITERATION
private static int	DEFAULT_MAX_NUM_EVALUATIONS_WOLF_SEARCH
private static int	DEFAULT_REPORT_EVERY
private static int	DEFAULT_RESTART_EVERY
private static double	DEFAULT_TOLERANCE
private double	extendAlphaFactorWolfSearch
protected double[]	G
private static double	initStepSteepestDecent
private int	iterationNum
protected LineSearch	lineSearch
private double	magnitudeForce
private int	maxNumEvaluationsWolfSearch
private static int	maxSteepestDecent
private int	nSteepestDecent
private static int	numStepsSteepestDecent
protected double[]	P
protected int	restartEvery
private static double	stepSizeExpansionSteepestDecent
private static double	stepSizeReductionSteepestDecent

Fields inherited from class meshi.optimizers.Minimizer

MAX_KICKSTARTS, terminator, tolerance

Fields inherited from class meshi.optimizers.Optimizer

energy, maxSteps, optimizerTerminator, reportEvery

Constructor Summary

ConjugateGradient(TotalEnergy energy)

ConjugateGradient(TotalEnergy energy, double tolerance, int maxSteps, int reportEvery)

ConjugateGradient(TotalEnergy energy, double tolerance, int maxSteps, int reportEvery, double c1, double c2, double extendAlphaFactorWolfSearch, int maxNumEvaluationsWolfSearch, int restartEvery)

ConjugateGradient(TotalEnergy energy, double tolerance, int maxSteps, int reportEvery, int restartEvery)

Method Summary

protected void	init()
protected void	kickStart()
protected boolean	minimizationStep()
java.lang.String	toString()

Methods inherited from class meshi.optimizers.Minimizer

run, run

Methods inherited from class meshi.optimizers.Optimizer

energy

Methods inherited from class java.lang.Object

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

Field Detail

lineSearch

protected LineSearch lineSearch

coordinates

protected double[][] coordinates

bufferCoordinates

protected double[][] bufferCoordinates

P

protected double[] P

G

protected double[] G

beta

protected double beta

DEFAULT_RESTART_EVERY

private static final int DEFAULT_RESTART_EVERY

See Also:

Constant Field Values

restartEvery

protected int restartEvery

iterationNum

private int iterationNum

magnitudeForce

private double magnitudeForce

DEFAULT_TOLERANCE

private static final double DEFAULT_TOLERANCE

See Also:

Constant Field Values

DEFAULT_MAX_ITERATION

private static final int DEFAULT_MAX_ITERATION

See Also:

Constant Field Values

DEFAULT_REPORT_EVERY

private static final int DEFAULT_REPORT_EVERY

See Also:

Constant Field Values

initStepSteepestDecent

private static double initStepSteepestDecent

stepSizeReductionSteepestDecent

private static double stepSizeReductionSteepestDecent

stepSizeExpansionSteepestDecent

private static double stepSizeExpansionSteepestDecent

numStepsSteepestDecent

private static int numStepsSteepestDecent

maxSteepestDecent

private static int maxSteepestDecent

nSteepestDecent

private int nSteepestDecent

c1

private double c1

c2

private double c2

extendAlphaFactorWolfSearch

private double extendAlphaFactorWolfSearch

maxNumEvaluationsWolfSearch

private int maxNumEvaluationsWolfSearch

DEFAULT_C1

private static final double DEFAULT_C1

See Also:

Constant Field Values

DEFAULT_C2

private static final double DEFAULT_C2

See Also:

Constant Field Values

DEFAULT_EXTENDED_ALPHA_FACTOR_WOLF_SEARCH

private static final double DEFAULT_EXTENDED_ALPHA_FACTOR_WOLF_SEARCH

See Also:

Constant Field Values

DEFAULT_MAX_NUM_EVALUATIONS_WOLF_SEARCH

private static final int DEFAULT_MAX_NUM_EVALUATIONS_WOLF_SEARCH

See Also:

Constant Field Values

Constructor Detail

ConjugateGradient

public ConjugateGradient(TotalEnergy energy,
                         double tolerance,
                         int maxSteps,
                         int reportEvery,
                         double c1,
                         double c2,
                         double extendAlphaFactorWolfSearch,
                         int maxNumEvaluationsWolfSearch,
                         int restartEvery)

ConjugateGradient

public ConjugateGradient(TotalEnergy energy,
                         double tolerance,
                         int maxSteps,
                         int reportEvery)

ConjugateGradient

public ConjugateGradient(TotalEnergy energy,
                         double tolerance,
                         int maxSteps,
                         int reportEvery,
                         int restartEvery)

ConjugateGradient

public ConjugateGradient(TotalEnergy energy)

Method Detail

init

protected void init()

Specified by:

init in class Minimizer

minimizationStep

protected boolean minimizationStep()
                            throws OptimizerException

Specified by:

minimizationStep in class Minimizer

Throws:

OptimizerException

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object

kickStart

protected void kickStart()

Specified by:

kickStart in class Minimizer