de.jungblut.clustering

Class MeanShiftClustering

java.lang.Object

de.jungblut.clustering.MeanShiftClustering

public final class MeanShiftClustering
extends Object

Sequential Mean Shift Clustering using a gaussian kernel and euclidian distance measurement.

Author:

thomasjungblut

Constructor Summary

Constructors

Constructor and Description
MeanShiftClustering()

Method Summary

Modifier and Type	Method and Description
static List<de.jungblut.math.DoubleVector>	cluster(List<de.jungblut.math.DoubleVector> points, double windowSize, double mergeWindow, int maxIterations, boolean verbose) Clusters a bunch of given points using the Mean Shift algorithm.

Methods inherited from class java.lang.Object

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

Constructor Detail

MeanShiftClustering

public MeanShiftClustering()

Method Detail

cluster

public static List<de.jungblut.math.DoubleVector> cluster(List<de.jungblut.math.DoubleVector> points,
                                                          double windowSize,
                                                          double mergeWindow,
                                                          int maxIterations,
                                                          boolean verbose)

Clusters a bunch of given points using the Mean Shift algorithm. It first observes possible centers that are within the given windowSize. Once we have initial centers found, we do a meanshift step and afterwards merge centers that are within the mergeWindow of each other. This algorithm is guranteed to converge to a minimum solution.

Parameters:

points - the points to cluster.

windowSize - the window size to observe points arround the center in. This is also used to observe initial centers.

mergeWindow - the merge window size, if a pair of centers is within this mergeWindow the centers are merged together.

maxIterations - the maximum number of iterations to do before breaking.

verbose - if true, progress will be reported after each iteration.

Returns:

the centers of the meanshift algorithm.