keel.Algorithms.Semi_Supervised_Learning.CLCC

Class CLCCGenerator

java.lang.Object

keel.Algorithms.Semi_Supervised_Learning.Basic.PrototypeGenerator

keel.Algorithms.Semi_Supervised_Learning.CLCC.CLCCGenerator

public class CLCCGenerator
extends PrototypeGenerator

This class implements the CLCC. 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.

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

Method Summary

Modifier and Type	Method and Description
Pair<PrototypeSet,PrototypeSet>	applyAlgorithm() Apply the CLCCGenerator method.
PrototypeSet	co_forest_sim(PrototypeSet labeled, PrototypeSet unlabeled) It applies a coforest-sim algorithm and fill the this.confidence matrix.
PrototypeSet[]	createCluster(PrototypeSet clusterCenters, PrototypeSet Lstar) Creates a cluster from the centers and the complete set of prototypes
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.
PrototypeSet[]	localClusterCenter(PrototypeSet Lstar) Returns the local cluster centers for a given prototype set.
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.
double	objetiveFunction(PrototypeSet[] cluster, PrototypeSet Lstar, PrototypeSet centers) Returns the objective function value of the clusters given.
double	penalty(int numCluster, int n) Returns the penalty for the cluster given.
PrototypeSet	ProcessCluster(PrototypeSet[] CX, PrototypeSet Lstar) Process the cluster given to obtain the best cluster evaluated with the data given.
PrototypeSet	resampleWithWeights(PrototypeSet data, int id, boolean[] sampled) Resample instances w.r.t the weight
int	votingRule(Prototype inst) Returns the predicted class for a 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.

numberOfClass

protected int numberOfClass

Number of classes.

Constructor Detail

CLCCGenerator

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

Build a new CLCCGenerator Algorithm

Parameters:

_trainingDataSet - Original prototype set to be reduced.

neigbors - number of neighbours considered. (not used)

poblacion - population size. (not used)

perc - Reduction percentage of the prototype set.

iteraciones - number of iterations. (not used)

wend - ending w value. (not used)

c1 - class 1 value. (not used)

vmax - maximum v value. (not used)

c2 - class 2 value. (not used)

wstart - starting w value. (not used)

CLCCGenerator

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

Build a new CLCCGenerator Algorithm

Parameters:

t - Original prototype set to be reduced.

unlabeled - Original unlabeled prototype set for SSL.

test - Origital 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

Returns the predicted class for a given prototype.

Parameters:

inst - given prototype.

Returns:

the predicted class for a given prototype.

Throws:

java.lang.Exception

co_forest_sim

public PrototypeSet co_forest_sim(PrototypeSet labeled,
                                  PrototypeSet unlabeled)
                           throws java.lang.Exception

It applies a coforest-sim algorithm and fill the this.confidence matrix.

Parameters:

labeled -

unlabeled -

Returns:

Throws:

java.lang.Exception

penalty

public double penalty(int numCluster,
                      int n)

Returns the penalty for the cluster given.

Parameters:

numCluster - number of clusters

n - number of instances.

Returns:

the penalty for the cluster given.

objetiveFunction

public double objetiveFunction(PrototypeSet[] cluster,
                               PrototypeSet Lstar,
                               PrototypeSet centers)

Returns the objective function value of the clusters given.

Parameters:

cluster - vector that codes the clusters.

Lstar - prototype set to evaluate the clusters.

centers - center of each cluster.

Returns:

the objective function value of the clusters given.

createCluster

public PrototypeSet[] createCluster(PrototypeSet clusterCenters,
                                    PrototypeSet Lstar)

Creates a cluster from the centers and the complete set of prototypes

Parameters:

clusterCenters - cluster centers given.

Lstar - complete prototype set.

Returns:

the clusters.

localClusterCenter

public PrototypeSet[] localClusterCenter(PrototypeSet Lstar)

Returns the local cluster centers for a given prototype set.

Parameters:

Lstar - given prototype set.

Returns:

the local cluster centers for a given prototype set.

ProcessCluster

public PrototypeSet ProcessCluster(PrototypeSet[] CX,
                                   PrototypeSet Lstar)

Process the cluster given to obtain the best cluster evaluated with the data given.

Parameters:

CX - Clusters centers.

Lstar - prototype set to evaluate the different clusters.

Returns:

Prototypes belonging to the best cluster.

applyAlgorithm

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

Apply the CLCCGenerator method.

Overrides:

applyAlgorithm in class PrototypeGenerator

Returns:

transductive test sets.

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.