Package elki.clustering.kmeans

Class BetulaLloydKMeans

java.lang.Object

elki.clustering.kmeans.AbstractKMeans<elki.data.NumberVector,KMeansModel>

elki.clustering.kmeans.BetulaLloydKMeans

All Implemented Interfaces:

elki.Algorithm, ClusteringAlgorithm<Clustering<KMeansModel>>, KMeans<elki.data.NumberVector,KMeansModel>

@Reference(authors="Andreas Lang and Erich Schubert",
           title="BETULA: Fast Clustering of Large Data with Improved BIRCH CF-Trees",
           booktitle="Information Systems",
           url="https://doi.org/10.1016/j.is.2021.101918",
           bibkey="DBLP:journals/is/LangS22")
public class BetulaLloydKMeans
extends AbstractKMeans<elki.data.NumberVector,KMeansModel>

BIRCH/BETULA-based clustering algorithm that simply treats the leafs of the CFTree as clusters.

References:

Andreas Lang and Erich Schubert
BETULA: Fast Clustering of Large Data with Improved BIRCH CF-Trees
Information Systems

Since:

0.8.0

Author:

Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	BetulaLloydKMeans.Par	Parameterization class.

Nested classes/interfaces inherited from class elki.clustering.kmeans.AbstractKMeans

AbstractKMeans.Instance

Nested classes/interfaces inherited from interface elki.Algorithm

elki.Algorithm.Utils

Field Summary

Fields

Modifier and Type	Field	Description
(package private) CFTree.Factory<?>	cffactory	CFTree factory.
(package private) long	diststat	Number of distance caclulations
(package private) boolean	ignoreWeight	Ignore weight
(package private) AbstractCFKMeansInitialization	initialization	k-means++ initialization
private static elki.logging.Logging	LOG	Class logger.
(package private) boolean	storeIds	Store ids

Fields inherited from class elki.clustering.kmeans.AbstractKMeans

distance, initializer, k, maxiter

Fields inherited from interface elki.clustering.kmeans.KMeans

DISTANCE_FUNCTION_ID, INIT_ID, K_ID, MAXITER_ID, SEED_ID, VARSTAT_ID

Constructor Summary

Constructors

Constructor	Description
BetulaLloydKMeans(int k, int maxiter, CFTree.Factory<?> cffactory, AbstractCFKMeansInitialization initialization, boolean storeIds, boolean ignoreWeight)	Constructor.

Method Summary

Modifier and Type	Method	Description
private int	assignToNearestCluster(int[] assignment, double[][] means, java.util.ArrayList<? extends ClusterFeature> cfs, int[] weights)	Assign each element to nearest cluster.
protected double[]	calculateVariances(int[] assignment, double[][] means, java.util.ArrayList<? extends ClusterFeature> cfs, int[] weights)	Calculate variance of clusters based on clustering features.
private double	distance(double[] x, double[] y)	Updates statistics and calculates distance between two Objects based on selected criteria.
private double	distance(elki.data.NumberVector x, double[] y)	Updates statistics and calculates distance between two Objects based on selected criteria.
protected elki.logging.Logging	getLogger()	Get the (STATIC) logger for this class.
private double[][]	kmeans(java.util.ArrayList<? extends ClusterFeature> cfs, int[] assignment, int[] weights, CFTree<?> tree)	Perform k-means clustering.
private double[][]	means(int[] assignment, double[][] means, java.util.ArrayList<? extends ClusterFeature> cfs, int[] weights)	Calculate means of clusters.
Clustering<KMeansModel>	run(elki.database.relation.Relation<elki.data.NumberVector> relation)	Run the clustering algorithm.

Methods inherited from class elki.clustering.kmeans.AbstractKMeans

getDistance, getInputTypeRestriction, incrementalUpdateMean, initialMeans, means, minusEquals, nearestMeans, plusEquals, plusMinusEquals, setDistance, setInitializer, setK

Methods inherited from class java.lang.Object

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

Methods inherited from interface elki.clustering.ClusteringAlgorithm

autorun

Field Detail

LOG

private static final elki.logging.Logging LOG

Class logger.

cffactory

CFTree.Factory<?> cffactory

CFTree factory.

initialization

AbstractCFKMeansInitialization initialization

k-means++ initialization

storeIds

boolean storeIds

Store ids

ignoreWeight

boolean ignoreWeight

Ignore weight

diststat

long diststat

Number of distance caclulations

Constructor Detail

BetulaLloydKMeans

public BetulaLloydKMeans(int k,
                         int maxiter,
                         CFTree.Factory<?> cffactory,
                         AbstractCFKMeansInitialization initialization,
                         boolean storeIds,
                         boolean ignoreWeight)

Constructor.

Parameters:

k - Number of clusters

maxiter - Maximum number of iterations

cffactory - CFTree factory

initialization - Initialization method for k-means

storeIds - Store IDs to avoid reassignment cost

ignoreWeight - Ignore the leaf weights

Method Detail

run

public Clustering<KMeansModel> run(elki.database.relation.Relation<elki.data.NumberVector> relation)

Run the clustering algorithm.

Parameters:

relation - Input data

Returns:

Clustering

kmeans

private double[][] kmeans(java.util.ArrayList<? extends ClusterFeature> cfs,
                          int[] assignment,
                          int[] weights,
                          CFTree<?> tree)

Perform k-means clustering.

Parameters:

cfs - Cluster features

assignment - Cluster assignment of each CF

weights - Cluster weight output

tree - CF tree

Returns:

Cluster means

means

private double[][] means(int[] assignment,
                         double[][] means,
                         java.util.ArrayList<? extends ClusterFeature> cfs,
                         int[] weights)

Calculate means of clusters.

Parameters:

assignment - Cluster assignment

means - Means of clusters

cfs - Clustering features

weights - Cluster weights

Returns:

Means of clusters.

assignToNearestCluster

private int assignToNearestCluster(int[] assignment,
                                   double[][] means,
                                   java.util.ArrayList<? extends ClusterFeature> cfs,
                                   int[] weights)

Assign each element to nearest cluster.

Parameters:

assignment - Current cluster assignment

means - k-means cluster means

cfs - Cluster features

weights - Cluster weights (output)

Returns:

Number of reassigned elements

calculateVariances

protected double[] calculateVariances(int[] assignment,
                                      double[][] means,
                                      java.util.ArrayList<? extends ClusterFeature> cfs,
                                      int[] weights)

Calculate variance of clusters based on clustering features.

The result is only correct after updating the means!

Parameters:

assignment - Cluster assignment of CFs

means - Cluster means

cfs - CF leaves

weights - Cluster weights

Returns:

Per-cluster variances

distance

private double distance(elki.data.NumberVector x,
                        double[] y)

Updates statistics and calculates distance between two Objects based on selected criteria.

Note: specializing this rather than calling SquaredEuclideanDistance was much faster, as we can avoid wrapping the array.

Parameters:

x - Point x

y - Point y

Returns:

distance

distance

private double distance(double[] x,
                        double[] y)

Updates statistics and calculates distance between two Objects based on selected criteria.

Parameters:

x - Point x

y - Point y

Returns:

distance

getLogger

protected elki.logging.Logging getLogger()

Description copied from class: AbstractKMeans

Get the (STATIC) logger for this class.

Specified by:

getLogger in class AbstractKMeans<elki.data.NumberVector,KMeansModel>

Returns:

the static logger