hex.optimization

Class OptimizationUtils.ExactLineSearch

java.lang.Object

hex.optimization.OptimizationUtils.ExactLineSearch

Enclosing class:

OptimizationUtils

public static final class OptimizationUtils.ExactLineSearch
extends java.lang.Object

This class implements the exact line search described in the doc, Algorithm 11.5

Field Summary

Fields

Modifier and Type	Field and Description
double	_alphai
double	_alphal
double	_alphar
double	_betaLS1
double	_betaLS2
java.lang.String[]	_coeffNames
double	_currGradDirIP
double[]	_direction
GLM.GLMGradientInfo	_ginfoOriginal
double	_lambdaLS
int	_maxIteration
double[]	_newBeta
double[]	_originalBeta

Constructor Summary

Constructors

Constructor and Description
ExactLineSearch(double[] betaCnd, ComputationState state, java.util.List<java.lang.String> coeffNames)

Method Summary

Modifier and Type	Method and Description
boolean	evaluateFirstWolfe(GLM.GLMGradientInfo ginfoNew) Evaluate and make sure that step size alphi is not too big so that objective function is still decreasing.
boolean	evaluateSecondWolfe(GLM.GLMGradientInfo ginfo) Evaluate and make sure that step size alphi is not too small so that good progress is made in reducing the loss function.
boolean	findAlpha(double[] lambdaEqual, double[] lambdaLessThan, ComputationState state, ConstrainedGLMUtils.LinearConstraints[] equalityConstraints, ConstrainedGLMUtils.LinearConstraints[] lessThanEqualToConstraints, GLM.GLMGradientSolver gradientSolver) Implements the Line Search algorithm in the doc, Algorithm 11.5.
void	reset(double[] betaCnd, ComputationState state, java.util.List<java.lang.String> coeffNames)
boolean	setAlphai(boolean firstWolfe, boolean secondWolfe)
void	setBetaConstraintsDeriv(double[] lambdaEqual, double[] lambdaLessThan, ComputationState state, ConstrainedGLMUtils.LinearConstraints[] equalityConstraints, ConstrainedGLMUtils.LinearConstraints[] lessThanEqualToConstraints, GLM.GLMGradientSolver gradientSolver, double[] betaCnd)

Methods inherited from class java.lang.Object

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

Field Detail

_betaLS1

public final double _betaLS1

See Also:

Constant Field Values

_betaLS2

public final double _betaLS2

See Also:

Constant Field Values

_lambdaLS

public final double _lambdaLS

See Also:

Constant Field Values

_alphal

public double _alphal

_alphar

public double _alphar

_alphai

public double _alphai

_direction

public double[] _direction

_maxIteration

public int _maxIteration

_originalBeta

public double[] _originalBeta

_newBeta

public double[] _newBeta

_ginfoOriginal

public GLM.GLMGradientInfo _ginfoOriginal

_currGradDirIP

public double _currGradDirIP

_coeffNames

public java.lang.String[] _coeffNames

Constructor Detail

ExactLineSearch

public ExactLineSearch(double[] betaCnd,
                       ComputationState state,
                       java.util.List<java.lang.String> coeffNames)

Method Detail

reset

public void reset(double[] betaCnd,
                  ComputationState state,
                  java.util.List<java.lang.String> coeffNames)

evaluateFirstWolfe

public boolean evaluateFirstWolfe(GLM.GLMGradientInfo ginfoNew)

Evaluate and make sure that step size alphi is not too big so that objective function is still decreasing. Refer to the doc, Definition 11.6

evaluateSecondWolfe

public boolean evaluateSecondWolfe(GLM.GLMGradientInfo ginfo)

Evaluate and make sure that step size alphi is not too small so that good progress is made in reducing the loss function. Refer to the doc, Definition 11.8

setAlphai

public boolean setAlphai(boolean firstWolfe,
                         boolean secondWolfe)

setBetaConstraintsDeriv

public void setBetaConstraintsDeriv(double[] lambdaEqual,
                                    double[] lambdaLessThan,
                                    ComputationState state,
                                    ConstrainedGLMUtils.LinearConstraints[] equalityConstraints,
                                    ConstrainedGLMUtils.LinearConstraints[] lessThanEqualToConstraints,
                                    GLM.GLMGradientSolver gradientSolver,
                                    double[] betaCnd)

findAlpha

public boolean findAlpha(double[] lambdaEqual,
                         double[] lambdaLessThan,
                         ComputationState state,
                         ConstrainedGLMUtils.LinearConstraints[] equalityConstraints,
                         ConstrainedGLMUtils.LinearConstraints[] lessThanEqualToConstraints,
                         GLM.GLMGradientSolver gradientSolver)

Implements the Line Search algorithm in the doc, Algorithm 11.5.