mulan.data
Class GreedyLabelClustering

java.lang.Object
  mulan.data.GreedyLabelClustering

All Implemented Interfaces:

Serializable, LabelClustering

public class GreedyLabelClustering
extends Object
implements LabelClustering, Serializable

A class for clustering dependent label pairs into disjoint subsets.

The type of the learned dependencies is determined by the LabelPairsDependenceIdentifier supplied to the class constructor. The clustering process is straightforward: initially all labels are assumed to be independent. Then we start group the label pairs according to their dependence score from most to least dependent. An SubsetLearner is build for each new partition and its accuracy is evaluated in terms of the measure. The process of grouping labels continues as long as the accuracy improves (or at least is not reduced). A number of steps specified by allowedNonImprovementSteps without seeking any concomitant improvement in the accuracy is allowed. Such a �non-useful� partitions are filtered out and the algorithm continues to evaluate subsequent pairs of dependent labels until one of the stop conditions is reached. The possible stop conditions are:
- no more label pairs to consider;
- all labels are clustered into one single group;
- pair dependence score is below the specified criticalValue;
- the number of allowedNonImprovementSteps is exceeded.

Version:

05.05.2011

Author:

Lena Chekina (lenat@bgu.ac.il)

See Also:

Serialized Form

Constructor Summary

GreedyLabelClustering(MultiLabelLearner aMultiLabelLearner, Classifier aSingleLabelLearner, LabelPairsDependenceIdentifier dependenceIdentifier)
Initialize the GreedyLabelClustering with multilabel and single label learners and a method for labels dependence identification.

Method Summary

int[][]	determineClusters(MultiLabelInstances trainingSet) Determines labels partitioning into dependent sets.
int	getAllowedNonImprovementSteps()
double	getCriticalValue()
Measure	getMeasure()
MultiLabelLearner	getMultiLabelLearner()
int	getNumFolds()
Classifier	getSingleLabelLearner()
boolean	isInternalSubsetLearnerDebug()
boolean	isUseSubsetLearnerCache()
static String	partitionToString(int[][] partition) Returns a string representation of the labels partition.
void	setAllowedNonImprovementSteps(int allowedNonImprovementSteps)
void	setCriticalValue(double criticalValue)
void	setInternalSubsetLearnerDebug(boolean internalSubsetLearnerDebug)
void	setMeasure(Measure measure)
void	setNumFolds(int numFolds)
void	setUseSubsetLearnerCache(boolean useSubsetLearnerCache)

Methods inherited from class java.lang.Object

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

Constructor Detail

GreedyLabelClustering

public GreedyLabelClustering(MultiLabelLearner aMultiLabelLearner,
                             Classifier aSingleLabelLearner,
                             LabelPairsDependenceIdentifier dependenceIdentifier)

Initialize the GreedyLabelClustering with multilabel and single label learners and a method for labels dependence identification.

Parameters:

aMultiLabelLearner - - a learner for multilabel classification

aSingleLabelLearner - - a learner for single label classification

dependenceIdentifier - - a method for label pairs dependence identification

Method Detail

determineClusters

public int[][] determineClusters(MultiLabelInstances trainingSet)

Determines labels partitioning into dependent sets. It clusters label pairs according to their dependence score and evaluates the related models. The clustering process continues as long as the accuracy improves. The finally selected labels partition is returned.

Specified by:

determineClusters in interface LabelClustering

Parameters:

trainingSet - the training data set

Returns:

a label set partitioning

partitionToString

public static String partitionToString(int[][] partition)

Returns a string representation of the labels partition.

Parameters:

partition - - a label set partition

Returns:

a string representation of the labels partition

getNumFolds

public int getNumFolds()

Returns:

setNumFolds

public void setNumFolds(int numFolds)

Parameters:

numFolds -

getMeasure

public Measure getMeasure()

Returns:

setMeasure

public void setMeasure(Measure measure)

Parameters:

measure -

getAllowedNonImprovementSteps

public int getAllowedNonImprovementSteps()

Returns:

setAllowedNonImprovementSteps

public void setAllowedNonImprovementSteps(int allowedNonImprovementSteps)

Parameters:

allowedNonImprovementSteps -

getCriticalValue

public double getCriticalValue()

Returns:

setCriticalValue

public void setCriticalValue(double criticalValue)

Parameters:

criticalValue -

getSingleLabelLearner

public Classifier getSingleLabelLearner()

Returns:

getMultiLabelLearner

public MultiLabelLearner getMultiLabelLearner()

Returns:

isUseSubsetLearnerCache

public boolean isUseSubsetLearnerCache()

Returns:

setUseSubsetLearnerCache

public void setUseSubsetLearnerCache(boolean useSubsetLearnerCache)

Parameters:

useSubsetLearnerCache -

isInternalSubsetLearnerDebug

public boolean isInternalSubsetLearnerDebug()

Returns:

setInternalSubsetLearnerDebug

public void setInternalSubsetLearnerDebug(boolean internalSubsetLearnerDebug)

Parameters:

internalSubsetLearnerDebug -