de.jungblut.clustering

Class KMeansClustering

java.lang.Object

de.jungblut.clustering.KMeansClustering

public final class KMeansClustering
extends Object

Sequential version of k-means clustering.

Author:

thomas.jungblut

Constructor Summary

Constructors

Constructor and Description
KMeansClustering(int k, de.jungblut.math.DoubleVector[] vectors, boolean random) Initializes a new KMeansClustering.
KMeansClustering(int k, List<de.jungblut.math.DoubleVector> vectors, boolean random) Initializes a new KMeansClustering.
KMeansClustering(List<de.jungblut.math.DoubleVector> centers, List<de.jungblut.math.DoubleVector> vectors) Initializes a new KMeansClustering.

Method Summary

Modifier and Type	Method and Description
List<Cluster>	cluster(int iterations, DistanceMeasurer distanceMeasurer, double delta, boolean verbose) Starts the clustering process.
de.jungblut.math.DoubleVector[]	getCenters()
double	getClusteringCost()

Methods inherited from class java.lang.Object

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

Constructor Detail

KMeansClustering

public KMeansClustering(int k,
                        de.jungblut.math.DoubleVector[] vectors,
                        boolean random)

Initializes a new KMeansClustering.

Parameters:

k - the number of centers to use.

vectors - the vectors to cluster.

random - true if use random initialization, else it will just pick the first k vectors.

KMeansClustering

public KMeansClustering(int k,
                        List<de.jungblut.math.DoubleVector> vectors,
                        boolean random)

Initializes a new KMeansClustering.

Parameters:

k - the number of centers to use.

vectors - the vectors to cluster.

random - true if use random initialization, else it will just pick the first k vectors.

KMeansClustering

public KMeansClustering(List<de.jungblut.math.DoubleVector> centers,
                        List<de.jungblut.math.DoubleVector> vectors)

Initializes a new KMeansClustering.

Parameters:

centers - initial centers, maybe seeded from CanopyClustering.

vectors - the vectors to cluster.

Method Detail

cluster

public List<Cluster> cluster(int iterations,
                             DistanceMeasurer distanceMeasurer,
                             double delta,
                             boolean verbose)

Starts the clustering process.

Parameters:

iterations - the iterations to cluster.

distanceMeasurer - the distance measurement to use.

delta - is the change in the sum of distances over iterations. If the difference is lower than delta the iteration will stop.

if - true, costs in each iteration will be printed.

Returns:

the clusters, which contain a center and the assigned vectors.

getClusteringCost

public double getClusteringCost()

getCenters

public de.jungblut.math.DoubleVector[] getCenters()

Returns:

the current state of the centers.