Package elki.clustering.kmeans.initialization.betula

Class CFKPlusPlusTree

java.lang.Object

elki.clustering.kmeans.initialization.betula.AbstractCFKMeansInitialization

elki.clustering.kmeans.initialization.betula.CFKPlusPlusTree

@Alias("tree")
@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 CFKPlusPlusTree
extends AbstractCFKMeansInitialization

Initialize K-means by following tree paths weighted by their variance contribution. This is the strategy denoted "tree" in the reference.

References:

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

Since:

0.8.0

Author:

Andreas Lang

Nested Class Summary

Nested Classes

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

Field Summary

Fields

Modifier and Type	Field	Description
(package private) CFInitWeight	dist	Distance function to use for initial means
(package private) boolean	firstUniform	Choose the first center uniformly from the cluster features.
(package private) int	maxdepth	Maximum depth to choose at.

Fields inherited from class elki.clustering.kmeans.initialization.betula.AbstractCFKMeansInitialization

rf

Constructor Summary

Constructors

Constructor	Description
CFKPlusPlusTree(CFInitWeight dist, boolean firstUniform, int maxdepth, RandomFactory rf)	Constructor.

Method Summary

Modifier and Type	Method	Description
double[][]	chooseInitialMeans(CFTree<?> tree, java.util.List<? extends ClusterFeature> cfs, int k)	Build the initial models.
private AsClusterFeature	chooseNextNode(CFNode<?> current, java.util.List<? extends ClusterFeature> ccs, java.util.Random rnd)	Choose a child of the current node.

Methods inherited from class java.lang.Object

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

Field Detail

dist

CFInitWeight dist

Distance function to use for initial means

firstUniform

boolean firstUniform

Choose the first center uniformly from the cluster features.

maxdepth

int maxdepth

Maximum depth to choose at.

Constructor Detail

CFKPlusPlusTree

public CFKPlusPlusTree(CFInitWeight dist,
                       boolean firstUniform,
                       int maxdepth,
                       RandomFactory rf)

Constructor.

Parameters:

dist - distance function

firstUniform - choose first center uniformly from the leaves

maxdepth - maximum depth

rf - random generator

Method Detail

chooseInitialMeans

public double[][] chooseInitialMeans(CFTree<?> tree,
                                     java.util.List<? extends ClusterFeature> cfs,
                                     int k)

Description copied from class: AbstractCFKMeansInitialization

Build the initial models.

Specified by:

chooseInitialMeans in class AbstractCFKMeansInitialization

Parameters:

tree - CF tree

cfs - Cluster features of the tree (may be ignored for tree-based initializations, should be an array list for efficiency)

k - Number of clusters.

Returns:

initial cluster means

chooseNextNode

private AsClusterFeature chooseNextNode(CFNode<?> current,
                                        java.util.List<? extends ClusterFeature> ccs,
                                        java.util.Random rnd)

Choose a child of the current node.

Parameters:

current - Current node

ccs - Currently chosen cluster centers

rnd - Random generator

Returns:

New cluster center