moa.classifiers.functions

Class SPegasos

java.lang.Object

moa.AbstractMOAObject

moa.options.AbstractOptionHandler

moa.classifiers.AbstractClassifier

moa.classifiers.functions.SPegasos

All Implemented Interfaces:

Serializable, Classifier, AWTRenderable, MOAObject, OptionHandler

public class SPegasos
extends AbstractClassifier

Implements the stochastic variant of the Pegasos (Primal Estimated sub-GrAdient SOlver for SVM) method of Shalev-Shwartz et al. (2007). For more information, see

S. Shalev-Shwartz, Y. Singer, N. Srebro: Pegasos: Primal Estimated sub-GrAdient SOlver for SVM. In: 24th International Conference on MachineLearning, 807-814, 2007.

* BibTeX:

 @inproceedings{Shalev-Shwartz2007,
    author = {S. Shalev-Shwartz and Y. Singer and N. Srebro},
    booktitle = {24th International Conference on MachineLearning},
    pages = {807-814},
    title = {Pegasos: Primal Estimated sub-GrAdient SOlver for SVM},
    year = {2007}
 }
 

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected static int	HINGE
FloatOption	lambdaRegularizationOption
protected static int	LOGLOSS
MultiChoiceOption	lossFunctionOption
protected double	m_lambda The regularization parameter
protected int	m_loss The current loss function to minimize
protected double	m_t Holds the current iteration number
protected double[]	m_weights Stores the weights (+ bias in the last element)

Fields inherited from class moa.classifiers.AbstractClassifier

classifierRandom, modelContext, randomSeed, randomSeedOption, trainingWeightSeenByModel

Fields inherited from class moa.options.AbstractOptionHandler

classOptionNamesToPreparedObjects, options

Constructor Summary

Constructors

Constructor and Description
SPegasos()

Method Summary

Methods

Modifier and Type	Method and Description
protected double	dloss(double z)
protected static double	dotProd(weka.core.Instance inst1, double[] weights, int classIndex)
double	getLambda() Get the current value of lambda
int	getLossFunction() Get the current loss function.
void	getModelDescription(StringBuilder result, int indent) Returns a string representation of the model.
protected Measurement[]	getModelMeasurementsImpl() Gets the current measurements of this classifier. The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods.
String	getPurposeString() Gets the purpose of this object
double[]	getVotesForInstance(weka.core.Instance inst) Calculates the class membership probabilities for the given test instance.
boolean	isRandomizable() Gets whether this classifier needs a random seed.
void	reset() Reset the classifier.
void	resetLearningImpl() Resets this classifier.
void	setLambda(double lambda) Set the value of lambda to use
void	setLossFunction(int function) Set the loss function to use.
String	toString() Prints out the classifier.
void	trainOnInstanceImpl(weka.core.Instance instance) Trains the classifier with the given instance.

Methods inherited from class moa.classifiers.AbstractClassifier

contextIsCompatible, copy, correctlyClassifies, getAttributeNameString, getAWTRenderer, getClassLabelString, getClassNameString, getDescription, getModelContext, getModelMeasurements, getNominalValueString, getSubClassifiers, modelAttIndexToInstanceAttIndex, modelAttIndexToInstanceAttIndex, prepareForUseImpl, resetLearning, setModelContext, setRandomSeed, trainingHasStarted, trainingWeightSeenByModel, trainOnInstance

Methods inherited from class moa.options.AbstractOptionHandler

discoverOptionsViaReflection, getCLICreationString, getOptions, getPreparedClassOption, getPreparedClassOption, prepareClassOptions, prepareForUse, prepareForUse

Methods inherited from class moa.AbstractMOAObject

copy, measureByteSize, measureByteSize

Methods inherited from class java.lang.Object

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

Methods inherited from interface moa.options.OptionHandler

getCLICreationString, getOptions, prepareForUse, prepareForUse

Methods inherited from interface moa.MOAObject

measureByteSize

Field Detail

m_lambda

protected double m_lambda

The regularization parameter

lambdaRegularizationOption

public FloatOption lambdaRegularizationOption

HINGE

protected static final int HINGE

See Also:

Constant Field Values

LOGLOSS

protected static final int LOGLOSS

See Also:

Constant Field Values

m_loss

protected int m_loss

The current loss function to minimize

lossFunctionOption

public MultiChoiceOption lossFunctionOption

m_weights

protected double[] m_weights

Stores the weights (+ bias in the last element)

m_t

protected double m_t

Holds the current iteration number

Constructor Detail

SPegasos

public SPegasos()

Method Detail

getPurposeString

public String getPurposeString()

Description copied from interface: OptionHandler

Gets the purpose of this object

Specified by:

getPurposeString in interface OptionHandler

Overrides:

getPurposeString in class AbstractClassifier

Returns:

the string with the purpose of this object

setLambda

public void setLambda(double lambda)

Set the value of lambda to use

Parameters:

lambda - the value of lambda to use

getLambda

public double getLambda()

Get the current value of lambda

Returns:

the current value of lambda

setLossFunction

public void setLossFunction(int function)

Set the loss function to use.

Parameters:

function - the loss function to use.

getLossFunction

public int getLossFunction()

Get the current loss function.

Returns:

the current loss function.

reset

public void reset()

Reset the classifier.

dotProd

protected static double dotProd(weka.core.Instance inst1,
             double[] weights,
             int classIndex)

dloss

protected double dloss(double z)

resetLearningImpl

public void resetLearningImpl()

Description copied from class: AbstractClassifier

Resets this classifier. It must be similar to starting a new classifier from scratch.

The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods. Note that this will produce compiler errors if not overridden.

Specified by:

resetLearningImpl in class AbstractClassifier

trainOnInstanceImpl

public void trainOnInstanceImpl(weka.core.Instance instance)

Trains the classifier with the given instance.

Specified by:

trainOnInstanceImpl in class AbstractClassifier

Parameters:

instance - the new training instance to include in the model

getVotesForInstance

public double[] getVotesForInstance(weka.core.Instance inst)

Calculates the class membership probabilities for the given test instance.

Parameters:

instance - the instance to be classified

Returns:

predicted class probability distribution

getModelDescription

public void getModelDescription(StringBuilder result,
                       int indent)

Description copied from class: AbstractClassifier

Returns a string representation of the model.

Specified by:

getModelDescription in class AbstractClassifier

Parameters:

result - the stringbuilder to add the description

indent - the number of characters to indent

toString

public String toString()

Prints out the classifier.

Overrides:

toString in class AbstractMOAObject

Returns:

a description of the classifier as a string

getModelMeasurementsImpl

protected Measurement[] getModelMeasurementsImpl()

Description copied from class: AbstractClassifier

Gets the current measurements of this classifier.

The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods. Note that this will produce compiler errors if not overridden.

Specified by:

getModelMeasurementsImpl in class AbstractClassifier

Returns:

an array of measurements to be used in evaluation tasks

isRandomizable

public boolean isRandomizable()

Description copied from interface: Classifier

Gets whether this classifier needs a random seed. Examples of methods that needs a random seed are bagging and boosting.

Returns:

true if the classifier needs a random seed.