Package cc.mallet.classify

Class MaxEntOptimizableByLabelLikelihood

java.lang.Object

cc.mallet.classify.MaxEntOptimizableByLabelLikelihood

All Implemented Interfaces:

Optimizable, Optimizable.ByGradientValue

public class MaxEntOptimizableByLabelLikelihood
extends java.lang.Object
implements Optimizable.ByGradientValue

Nested Class Summary

Nested classes/interfaces inherited from interface cc.mallet.optimize.Optimizable

Optimizable.ByBatchGradient, Optimizable.ByCombiningBatchGradient, Optimizable.ByGISUpdate, Optimizable.ByGradient, Optimizable.ByGradientValue, Optimizable.ByHessian, Optimizable.ByValue, Optimizable.ByVotedPerceptron

Constructor Summary

Constructors

Constructor	Description
MaxEntOptimizableByLabelLikelihood()
MaxEntOptimizableByLabelLikelihood(InstanceList trainingSet, MaxEnt initialClassifier)

Method Summary

Modifier and Type	Method	Description
MaxEnt	getClassifier()
int	getNumParameters()
double	getParameter(int index)
void	getParameters(double[] buff)
double	getValue()
int	getValueCalls()	Counts how many times this trainer has computed the log probability of training labels.
void	getValueGradient(double[] buffer)
int	getValueGradientCalls()	Counts how many times this trainer has computed the gradient of the log probability of training labels.
MaxEntOptimizableByLabelLikelihood	setGaussianPriorVariance(double gaussianPriorVariance)	Sets a parameter to prevent overtraining.
MaxEntOptimizableByLabelLikelihood	setHyperbolicPriorSharpness(double hyperbolicPriorSharpness)
MaxEntOptimizableByLabelLikelihood	setHyperbolicPriorSlope(double hyperbolicPriorSlope)
void	setParameter(int index, double v)
void	setParameters(double[] buff)
MaxEntOptimizableByLabelLikelihood	useGaussianPrior()
MaxEntOptimizableByLabelLikelihood	useHyperbolicPrior()
MaxEntOptimizableByLabelLikelihood	useNoPrior()	In some cases a prior term is implemented in the optimizer, (eg orthant-wise L-BFGS), so we occasionally want to only calculate the log likelihood.

Methods inherited from class java.lang.Object

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

Constructor Detail

MaxEntOptimizableByLabelLikelihood

public MaxEntOptimizableByLabelLikelihood()

MaxEntOptimizableByLabelLikelihood

public MaxEntOptimizableByLabelLikelihood(InstanceList trainingSet,
                                          MaxEnt initialClassifier)

Method Detail

getClassifier

public MaxEnt getClassifier()

getParameter

public double getParameter(int index)

Specified by:

getParameter in interface Optimizable

setParameter

public void setParameter(int index,
                         double v)

Specified by:

setParameter in interface Optimizable

getNumParameters

public int getNumParameters()

Specified by:

getNumParameters in interface Optimizable

getParameters

public void getParameters(double[] buff)

Specified by:

getParameters in interface Optimizable

setParameters

public void setParameters(double[] buff)

Specified by:

setParameters in interface Optimizable

getValue

public double getValue()

Specified by:

getValue in interface Optimizable.ByGradientValue

getValueGradient

public void getValueGradient(double[] buffer)

Specified by:

getValueGradient in interface Optimizable.ByGradientValue

getValueGradientCalls

public int getValueGradientCalls()

Counts how many times this trainer has computed the gradient of the log probability of training labels.

getValueCalls

public int getValueCalls()

Counts how many times this trainer has computed the log probability of training labels.

useGaussianPrior

public MaxEntOptimizableByLabelLikelihood useGaussianPrior()

useHyperbolicPrior

public MaxEntOptimizableByLabelLikelihood useHyperbolicPrior()

useNoPrior

public MaxEntOptimizableByLabelLikelihood useNoPrior()

In some cases a prior term is implemented in the optimizer, (eg orthant-wise L-BFGS), so we occasionally want to only calculate the log likelihood.

setGaussianPriorVariance

public MaxEntOptimizableByLabelLikelihood setGaussianPriorVariance(double gaussianPriorVariance)

Sets a parameter to prevent overtraining. A smaller variance for the prior means that feature weights are expected to hover closer to 0, so extra evidence is required to set a higher weight.

Returns:

This trainer

setHyperbolicPriorSlope

public MaxEntOptimizableByLabelLikelihood setHyperbolicPriorSlope(double hyperbolicPriorSlope)

setHyperbolicPriorSharpness

public MaxEntOptimizableByLabelLikelihood setHyperbolicPriorSharpness(double hyperbolicPriorSharpness)