Package elki.clustering.kmeans.parallel

Class KMeansProcessor<V extends NumberVector>

java.lang.Object

elki.clustering.kmeans.parallel.KMeansProcessor<V>

All Implemented Interfaces:

Processor

public class KMeansProcessor<V extends NumberVector>
extends java.lang.Object
implements Processor

Parallel k-means implementation.

Since:

0.7.0

Author:

Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	KMeansProcessor.Instance<V extends NumberVector>	Instance to process part of the data set, for a single iteration.

Field Summary

Fields

Modifier and Type	Field	Description
(package private) WritableIntegerDataStore	assignment	Assignment storage.
(package private) double[][]	centroids	Updated cluster centroids
(package private) boolean	changed	Whether the assignment changed during the last iteration.
(package private) NumberVectorDistance<? super V>	distance	Distance function.
(package private) double[][]	means	Mean vectors.
(package private) Relation<V>	relation	Data relation.
(package private) int[]	sizes	(Partial) cluster sizes
(package private) double[]	varsum	Variance sum.

Constructor Summary

Constructors

Constructor	Description
KMeansProcessor(Relation<V> relation, NumberVectorDistance<? super V> distance, WritableIntegerDataStore assignment, double[] varsum)	Constructor.

Method Summary

Modifier and Type	Method	Description
boolean	changed()	Get the "has changed" value.
void	cleanup(Processor.Instance inst)	Invoke cleanup.
double[][]	getMeans()	Get the new means.
KMeansProcessor.Instance<V>	instantiate(Executor exectutor)	Create an instance.
void	nextIteration(double[][] means)	Initialize for a new iteration.

Methods inherited from class java.lang.Object

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

Field Detail

relation

Relation<V extends NumberVector> relation

Data relation.

distance

NumberVectorDistance<? super V extends NumberVector> distance

Distance function.

assignment

WritableIntegerDataStore assignment

Assignment storage.

means

double[][] means

Mean vectors.

centroids

double[][] centroids

Updated cluster centroids

sizes

int[] sizes

(Partial) cluster sizes

varsum

double[] varsum

Variance sum.

changed

boolean changed

Whether the assignment changed during the last iteration.

Constructor Detail

KMeansProcessor

public KMeansProcessor(Relation<V> relation,
                       NumberVectorDistance<? super V> distance,
                       WritableIntegerDataStore assignment,
                       double[] varsum)

Constructor.

Parameters:

relation - Data relation

distance - Distance function

assignment - Cluster assignment

varsum - Variance sums

Method Detail

changed

public boolean changed()

Get the "has changed" value.

Returns:

Changed flag.

nextIteration

public void nextIteration(double[][] means)

Initialize for a new iteration.

Parameters:

means - New means.

instantiate

public KMeansProcessor.Instance<V> instantiate(Executor exectutor)

Description copied from interface: Processor

Create an instance. May be called multiple times, for example for multiple threads.

Specified by:

instantiate in interface Processor

Parameters:

exectutor - Job executor

Returns:

Instance

cleanup

public void cleanup(Processor.Instance inst)

Description copied from interface: Processor

Invoke cleanup.

Specified by:

cleanup in interface Processor

Parameters:

inst - Instance to cleanup.

getMeans

public double[][] getMeans()

Get the new means.

Returns:

New means