de.jungblut.clustering

Class DBSCAN

java.lang.Object

de.jungblut.clustering.DBSCAN

public final class DBSCAN
extends Object

Sequential version of DBSCAN to evaluate if this algorithm is suitable for arbitrary parallelization paradigms that can crunch graphs.

PLAN:
1. compute distance matrix between the points
2. extract adjacent points via threshold epsilon and minpoints s
3. run connected components (here BFS)
4. PROFIT!

Constructor Summary

Constructors

Constructor and Description
DBSCAN()

Method Summary

Modifier and Type	Method and Description
ArrayList<de.jungblut.math.DoubleVector>[]	cluster(List<de.jungblut.math.DoubleVector> points, DistanceMeasurer measurer, int minPoints, double epsilon) Clusters the points.
List<de.jungblut.math.DoubleVector>	getNoise()

Methods inherited from class java.lang.Object

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

Constructor Detail

DBSCAN

public DBSCAN()

Method Detail

cluster

public ArrayList<de.jungblut.math.DoubleVector>[] cluster(List<de.jungblut.math.DoubleVector> points,
                                                          DistanceMeasurer measurer,
                                                          int minPoints,
                                                          double epsilon)

Clusters the points.

Parameters:

measurer - the distance measurer to use.

minPoints - the minimum points in a cluster.

epsilon - the radius of a point to detect other points.

getNoise

public List<de.jungblut.math.DoubleVector> getNoise()

Returns:

the found noise as list of vectors.