moa.cluster

Class SphereCluster

java.lang.Object

moa.AbstractMOAObject

moa.cluster.Cluster

moa.cluster.SphereCluster

All Implemented Interfaces:

Serializable, MOAObject

Direct Known Subclasses:

CFCluster

public class SphereCluster
extends Cluster

A simple implementation of the Cluster interface representing spherical clusters. The inclusion probability is one inside the sphere and zero everywhere else.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
SphereCluster()
SphereCluster(double[] center, double radius)
SphereCluster(double[] center, double radius, double weightedSize)
SphereCluster(int dimensions, double radius, Random random)
SphereCluster(List<? extends weka.core.Instance> instances, int dimension)

Method Summary

Methods

Modifier and Type	Method and Description
void	combine(SphereCluster cluster)
double[]	getCenter()
double	getCenterDistance(weka.core.Instance instance)
double	getCenterDistance(SphereCluster other)
protected void	getClusterSpecificInfo(ArrayList<String> infoTitle, ArrayList<String> infoValue)
double[]	getDistanceVector(weka.core.Instance instance)
double[]	getDistanceVector(SphereCluster other)
double	getHullDistance(SphereCluster other)
double	getInclusionProbability(weka.core.Instance instance) Returns the probability of the given point belonging to this cluster.
double	getRadius()
double	getWeight() Returns the weight of this cluster, not neccessarily normalized.
void	merge(SphereCluster cluster)
double	overlapRadiusDegree(SphereCluster other) Checks whether two SphereCluster overlap based on radius NOTE: overlapRadiusDegree only calculates the overlap based on the centers and the radi, so not the real overlap TODO: should we do this by MC to get the real overlap???
boolean	overlapSave(SphereCluster other) When a clusters looses points the new minimal bounding sphere can be partly outside of the originating cluster.
weka.core.Instance	sample(Random random) Samples this cluster by returning a point from inside it.
void	setCenter(double[] center)
void	setRadius(double radius)
void	setWeight(double weight)

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

Constructor Detail

SphereCluster

public SphereCluster(double[] center,
             double radius)

SphereCluster

public SphereCluster()

SphereCluster

public SphereCluster(double[] center,
             double radius,
             double weightedSize)

SphereCluster

public SphereCluster(int dimensions,
             double radius,
             Random random)

SphereCluster

public SphereCluster(List<? extends weka.core.Instance> instances,
             int dimension)

Method Detail

overlapRadiusDegree

public double overlapRadiusDegree(SphereCluster other)

Checks whether two SphereCluster overlap based on radius NOTE: overlapRadiusDegree only calculates the overlap based on the centers and the radi, so not the real overlap TODO: should we do this by MC to get the real overlap???

Parameters:

other -

Returns:

combine

public void combine(SphereCluster cluster)

merge

public void merge(SphereCluster cluster)

getCenter

public double[] getCenter()

Specified by:

getCenter in class Cluster

Returns:

the center of the cluster

setCenter

public void setCenter(double[] center)

getRadius

public double getRadius()

setRadius

public void setRadius(double radius)

getWeight

public double getWeight()

Description copied from class: Cluster

Returns the weight of this cluster, not neccessarily normalized. It could, for instance, simply return the number of points contined in this cluster.

Specified by:

getWeight in class Cluster

Returns:

the weight

setWeight

public void setWeight(double weight)

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 Cluster

Returns:

a value between 0 and 1

getCenterDistance

public double getCenterDistance(weka.core.Instance instance)

getCenterDistance

public double getCenterDistance(SphereCluster other)

getHullDistance

public double getHullDistance(SphereCluster other)

overlapSave

public boolean overlapSave(SphereCluster other)

When a clusters looses points the new minimal bounding sphere can be partly outside of the originating cluster. If a another cluster is right next to the original cluster (without overlapping), the new cluster can be overlapping with this second cluster. OverlapSave will tell you if the current cluster can degenerate so much that it overlaps with cluster 'other'

Parameters:

other - the potentially overlapping cluster

Returns:

true if cluster can potentially overlap

getDistanceVector

public double[] getDistanceVector(weka.core.Instance instance)

getDistanceVector

public double[] getDistanceVector(SphereCluster other)

sample

public weka.core.Instance sample(Random random)

Samples this cluster by returning a point from inside it.

Specified by:

sample in class Cluster

Parameters:

random - a random number source

Returns:

a point that lies inside this cluster

getClusterSpecificInfo

protected void getClusterSpecificInfo(ArrayList<String> infoTitle,
                          ArrayList<String> infoValue)

Overrides:

getClusterSpecificInfo in class Cluster