moa.clusterers.clustree

Class ClusKernel

java.lang.Object

moa.AbstractMOAObject

moa.cluster.Cluster

moa.cluster.SphereCluster

moa.cluster.CFCluster

moa.clusterers.clustree.ClusKernel

All Implemented Interfaces:

Serializable, MOAObject

public class ClusKernel
extends CFCluster

Representation of an Entry in the tree

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
static double	EPSILON Numeric epsilon.
static double	MIN_VARIANCE

Fields inherited from class moa.cluster.CFCluster

LS, N, radiusFactor, SS

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	ClusKernel(ClusKernel other) Instantiates a copy of the given cluster.
	ClusKernel(double[] point, int dim) A constructor that makes a Kernel which just represents the given point.
protected	ClusKernel(int numberDimensions) Constructor of the Cluster.

Method Summary

Methods

Modifier and Type	Method and Description
void	add(ClusKernel other) Adds the given cluster to this cluster, without making this cluster older.
protected void	aggregate(ClusKernel other, long timeDifference, double negLambda) Make this cluster older bei weighting it and add to this cluster the given cluster.
double	calcDistance(ClusKernel other) Calculate the distance to this other cluster.
protected void	clear() Remove all points from this cluster.
double[]	getCenter()
CFCluster	getCF()
double	getInclusionProbability(weka.core.Instance instance) Returns the probability of the given point belonging to this cluster.
double	getRadius() See interface Cluster
double[]	getVarianceVector()
double	getWeight() See interface Cluster
protected boolean	isEmpty() Check if this cluster is empty or not.
protected void	makeOlder(long timeDifference, double negLambda) Make this cluster older.
protected void	overwriteOldCluster(ClusKernel other) Overwrites the LS, SS and weightedN in this cluster to the values of the given cluster but adds N and classCount of the given cluster to this one.

Methods inherited from class moa.cluster.CFCluster

add, addVectors, getN, setN

Methods inherited from class moa.cluster.SphereCluster

combine, getCenterDistance, getCenterDistance, getClusterSpecificInfo, getDistanceVector, getDistanceVector, getHullDistance, merge, overlapRadiusDegree, overlapSave, sample, setCenter, setRadius, setWeight

Methods inherited from class moa.cluster.Cluster

getDescription, getGroundTruth, getId, getInfo, getMeasureValue, isGroundTruth, setGroundTruth, setId, setMeasureValue, setMeasureValue

Methods inherited from class moa.AbstractMOAObject

copy, copy, measureByteSize, measureByteSize, toString

Methods inherited from class java.lang.Object

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

Field Detail

EPSILON

public static final double EPSILON

Numeric epsilon.

See Also:

Constant Field Values

MIN_VARIANCE

public static final double MIN_VARIANCE

See Also:

Constant Field Values

Constructor Detail

ClusKernel

public ClusKernel(double[] point,
          int dim)

A constructor that makes a Kernel which just represents the given point.

Parameters:

point - The point to be converted into a corresponding Kernel.

numberClasses - The number of classes possible for points in this experiment(Tree).

ClusKernel

protected ClusKernel(int numberDimensions)

Constructor of the Cluster.

Parameters:

numberDimensions - Dimensionality of the points added that can be added to this cluster

numberClasses - The number of classes possible for points in this experiment(Tree).

ClusKernel

protected ClusKernel(ClusKernel other)

Instantiates a copy of the given cluster.

Parameters:

other - The Cluster of which we make a copy.

Method Detail

add

public void add(ClusKernel other)

Adds the given cluster to this cluster, without making this cluster older.

Parameters:

other -

aggregate

protected void aggregate(ClusKernel other,
             long timeDifference,
             double negLambda)

Make this cluster older bei weighting it and add to this cluster the given cluster. If we want to add somethin to the cluster, but don't want to weight it, we should use the function add(Cluster).

Parameters:

other - The other cluster to be added to this one.

timeDifference - The time elapsed between the last update of the Entry to which this cluster belongs and the update that caused the call to this function.

negLambda - A parameter needed to weight the cluster.

See Also:

#add(tree.Kernel)

makeOlder

protected void makeOlder(long timeDifference,
             double negLambda)

Make this cluster older. This means multiplying weighted N, LS and SS with a weight factor given by the time difference and the parameter negLambda.

Parameters:

timeDifference - The time elapsed between this current update and the last one.

negLambda -

calcDistance

public double calcDistance(ClusKernel other)

Calculate the distance to this other cluster. The other cluster is normaly just a single data point(i.e. N = 1).

Parameters:

other - The other cluster to which the distance is calculated.

Returns:

The distance between this cluster and the other.

isEmpty

protected boolean isEmpty()

Check if this cluster is empty or not.

Returns:

true if the cluster has no data points, false otherwise.

clear

protected void clear()

Remove all points from this cluster.

overwriteOldCluster

protected void overwriteOldCluster(ClusKernel other)

Overwrites the LS, SS and weightedN in this cluster to the values of the given cluster but adds N and classCount of the given cluster to this one. This function is useful when the weight of an entry becomes to small, and we want to forget the information of the old points.

Parameters:

other - The cluster that should overwrite the information.

getWeight

public double getWeight()

Description copied from class: CFCluster

See interface Cluster

Overrides:

getWeight in class CFCluster

Returns:

The weight.

See Also:

Cluster.getWeight()

getCF

public CFCluster getCF()

Specified by:

getCF in class CFCluster

getCenter

public double[] getCenter()

Overrides:

getCenter in class CFCluster

Returns:

this kernels' center

getInclusionProbability

public double getInclusionProbability(weka.core.Instance instance)

Description copied from class: Cluster

Returns the probability of the given point belonging to this cluster.

Specified by:

getInclusionProbability in class CFCluster

Returns:

a value between 0 and 1

getRadius

public double getRadius()

See interface Cluster

Specified by:

getRadius in class CFCluster

Returns:

The radius of the cluster.

See Also:

Cluster#getRadius()

getVarianceVector

public double[] getVarianceVector()