edu.ucla.sspace.clustering

Class CentroidCluster<T extends DoubleVector>

java.lang.Object

edu.ucla.sspace.clustering.CentroidCluster<T>

All Implemented Interfaces:

Cluster<T>

public class CentroidCluster<T extends DoubleVector>
extends Object
implements Cluster<T>

This Cluster represents only the centroid of the data points assigned to it. Assignments from data point indices to cluster ids are maintained internally, but the actualy data points are discareded.

Author:

Keith Stevens

Constructor Summary

Constructors

Constructor and Description
CentroidCluster(T emptyVector) Creates a new CentroidCluster that takes ownership of emptyVector as the centroid for this Cluster.

Method Summary

Methods

Modifier and Type	Method and Description
void	addVector(T vector, int id) Adds vector with id to this Cluster.
T	centroid() Returns an immutable view of the centroid for this Cluster.
double	compareWithVector(T vector) Returns the similarity between this Cluster and vector.
BitSet	dataPointIds() Returns the set of identifiers for all data points assigned to this cluster.
List<T>	dataPointValues() Returns an empty list, as no spare data point values are maintained.
void	merge(Cluster<T> other) Merges the other Cluster with this Cluster.
int	size() Returns the total number of points assigned to this Cluster.

Methods inherited from class java.lang.Object

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

Constructor Detail

CentroidCluster

public CentroidCluster(T emptyVector)

Creates a new CentroidCluster that takes ownership of emptyVector as the centroid for this Cluster. emptyVector should have length equal to the length of vectors that will be assigned to this Cluster and should be dense if a large number of vectors, or any dense vectors, are expected to be assigned to this Cluster.

Method Detail

addVector

public void addVector(T vector,
             int id)

Adds vector with id to this Cluster.

Specified by:

addVector in interface Cluster<T extends DoubleVector>

compareWithVector

public double compareWithVector(T vector)

Returns the similarity between this Cluster and vector.

Specified by:

compareWithVector in interface Cluster<T extends DoubleVector>

centroid

public T centroid()

Returns an immutable view of the centroid for this Cluster.

Specified by:

centroid in interface Cluster<T extends DoubleVector>

dataPointValues

public List<T> dataPointValues()

Returns an empty list, as no spare data point values are maintained.

Specified by:

dataPointValues in interface Cluster<T extends DoubleVector>

dataPointIds

public BitSet dataPointIds()

Returns the set of identifiers for all data points assigned to this cluster. TODO: Consider returning a set and let implementations return a wrapped BitSet, as in the graphical-update branch.

Specified by:

dataPointIds in interface Cluster<T extends DoubleVector>

merge

public void merge(Cluster<T> other)

Merges the other Cluster with this Cluster. This Cluster will absorb all data points from other and update the centroid and data point assignments as needed.

Specified by:

merge in interface Cluster<T extends DoubleVector>

size

public int size()

Returns the total number of points assigned to this Cluster.

Specified by:

size in interface Cluster<T extends DoubleVector>