keel.Algorithms.Semi_Supervised_Learning.ADE_CoForest

Class ADE_CoForestGenerator

java.lang.Object

keel.Algorithms.Semi_Supervised_Learning.Basic.PrototypeGenerator

keel.Algorithms.Semi_Supervised_Learning.ADE_CoForest.ADE_CoForestGenerator

public class ADE_CoForestGenerator
extends PrototypeGenerator

This class implements the Tri-training. You can use: Knn, C4.5, SMO and NB as classifiers.

Author:

triguero

Field Summary

Fields

Modifier and Type	Field and Description
protected int	m_KValue Final number of features that were considered in last build.
protected int	m_numFeatures Number of features to consider in random feature selection.
protected int	numberOfClass Number of classes.
protected int	numberOfPrototypes Number of prototypes to generated.

Fields inherited from class keel.Algorithms.Semi_Supervised_Learning.Basic.PrototypeGenerator

algorithmName, generatedDataSet, Instancetest, Instancetrain, SEED, seedDefaultValueList, testDataSet, trainingDataSet, transductiveDataSet

Constructor Summary

Constructors

Constructor and Description
ADE_CoForestGenerator(PrototypeSet _trainingDataSet, int neigbors, int poblacion, int perc, int iteraciones, double c1, double c2, double vmax, double wstart, double wend) Build a new ADE_CoForestGenerator Algorithm
ADE_CoForestGenerator(PrototypeSet t, PrototypeSet unlabeled, PrototypeSet test, Parameters parameters) Build a new ADE_CoForestGenerator Algorithm

Method Summary

Modifier and Type	Method and Description
Pair<PrototypeSet,PrototypeSet>	applyAlgorithm() Apply the ADE_CoForestGenerator method.
protected boolean	isHighConfidence(Prototype inst, int idExcluded) To judege whether the confidence for a given instance of H* is high enough, which is affected by the onfidence threshold.
static void	main(java.lang.String[] args) General main for all the prototoype generators Arguments: 0: Filename with the training data set to be condensed. 1: Filename which contains the test data set. 3: Seed of the random number generator.
protected PrototypeSet	removeOnly(PrototypeSet T, PrototypeSet labeled) Depuration algorithm.
PrototypeSet	resampleWithWeights(PrototypeSet data, int id, boolean[] sampled) Resample instances w.r.t the weight
int	votingRule(Prototype inst) Predicted class of the given prototype.

Methods inherited from class keel.Algorithms.Semi_Supervised_Learning.Basic.PrototypeGenerator

absoluteAccuracy, absoluteAccuracyAndError, absoluteAccuracyKNN, accuracy, accuracy2, desordenar_vector_sin, desordenar_vector, execute, generateReducedDataSet, getResultingAccuracy, getResultsOfAccuracy, getSeed, getSetSizeFromPercentage, getSetSizeFromPercentage, getTime, inic_vector_sin, inic_vector, saveResultsOfAccuracyIn, selecRandomSet, setInstanceTest, setInstanceTrain, setSeed

Methods inherited from class java.lang.Object

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

Field Detail

m_numFeatures

protected int m_numFeatures

Number of features to consider in random feature selection. If less than 1 will use int(logM+1) )

m_KValue

protected int m_KValue

Final number of features that were considered in last build.

numberOfPrototypes

protected int numberOfPrototypes

Number of prototypes to generated.

numberOfClass

protected int numberOfClass

Number of classes.

Constructor Detail

ADE_CoForestGenerator

public ADE_CoForestGenerator(PrototypeSet _trainingDataSet,
                             int neigbors,
                             int poblacion,
                             int perc,
                             int iteraciones,
                             double c1,
                             double c2,
                             double vmax,
                             double wstart,
                             double wend)

Build a new ADE_CoForestGenerator Algorithm

Parameters:

_trainingDataSet - Original prototype set to be reduced.

neigbors - Number of neighbours considered.

poblacion - population size.

perc - Reduction percentage of the prototype set.

iteraciones - number of iterations.

c2 - Class 2.

c1 - Class 1.

vmax - max value.

wend - ending value of w.

wstart - starting value of w.

ADE_CoForestGenerator

public ADE_CoForestGenerator(PrototypeSet t,
                             PrototypeSet unlabeled,
                             PrototypeSet test,
                             Parameters parameters)

Build a new ADE_CoForestGenerator Algorithm

Parameters:

t - Original prototype set to be reduced.

unlabeled - Original unlabeled prototype set for SSL.

test - Test prototype set.

parameters - Parameters of the algorithm (only % of reduced set).

Method Detail

resampleWithWeights

public final PrototypeSet resampleWithWeights(PrototypeSet data,
                                              int id,
                                              boolean[] sampled)

Resample instances w.r.t the weight

Parameters:

data - Instances -- the original data set

id - of the classifier

sampled - boolean[] -- the output parameter, indicating whether the instance is sampled

Returns:

Instances

isHighConfidence

protected boolean isHighConfidence(Prototype inst,
                                   int idExcluded)
                            throws java.lang.Exception

To judege whether the confidence for a given instance of H* is high enough, which is affected by the onfidence threshold. Meanwhile, if the example is the confident one, assign label to it and weigh the example with the confidence

Parameters:

inst - Instance -- The instance

idExcluded - int -- the index of the individual should be excluded from H*

Returns:

boolean -- true for high

Throws:

java.lang.Exception - - some exception

votingRule

public int votingRule(Prototype inst)
               throws java.lang.Exception

Predicted class of the given prototype.

Parameters:

inst - given prototype.

Returns:

predicted class of the given prototype.

Throws:

java.lang.Exception

removeOnly

protected PrototypeSet removeOnly(PrototypeSet T,
                                  PrototypeSet labeled)

Depuration algorithm. Filters the prototypes from the given set T.

Parameters:

T - set to be filtered.

labeled - labeled prototypes ued to filter the rest.

Returns:

Filtered Prototypes.

applyAlgorithm

public Pair<PrototypeSet,PrototypeSet> applyAlgorithm()
                                               throws java.lang.Exception

Apply the ADE_CoForestGenerator method.

Overrides:

applyAlgorithm in class PrototypeGenerator

Returns:

Pair of prototypes set (transductive, test)

Throws:

java.lang.Exception - if the algorithm can not be applied.

main

public static void main(java.lang.String[] args)

General main for all the prototoype generators Arguments: 0: Filename with the training data set to be condensed. 1: Filename which contains the test data set. 3: Seed of the random number generator. Always. **************************

Parameters:

args - Arguments of the main function.