de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.extraction

Class AbstractCutDendrogram.Instance

java.lang.Object

de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.extraction.AbstractCutDendrogram.Instance

Direct Known Subclasses:

CutDendrogramByHeight.Instance, CutDendrogramByNumberOfClusters.Instance

Enclosing class:

AbstractCutDendrogram

public abstract class AbstractCutDendrogram.Instance
extends java.lang.Object

Instance for a single data set.

Author:

Erich Schubert

Field Summary

Fields

Modifier and Type	Field and Description
protected java.util.ArrayList<ModifiableDBIDs>	cluster_dbids Storage for cluster contents
protected ArrayModifiableDBIDs	cluster_leads Cluster lead objects
protected WritableIntegerDataStore	cluster_map Map clusters to integer cluster numbers.
protected DoubleArray	clusterHeight Cluster distances
protected ArrayDBIDs	ids Unordered IDs
protected DoubleDataStore	lambda Merge distance
protected DBIDDataStore	pi Parent pointer
protected PointerHierarchyRepresentationResult	pointerresult The hierarchical result to process.

Constructor Summary

Constructors

Constructor and Description
Instance(PointerHierarchyRepresentationResult pointerresult) Constructor.

Method Summary

Modifier and Type	Method and Description
private Clustering<DendrogramModel>	buildFlat(DBIDArrayIter it, int split, FiniteProgress progress) Build a flat clustering.
private Clustering<DendrogramModel>	buildHierarchical(DBIDArrayIter it, int split, FiniteProgress progress) Build a hierarchical clustering.
private void	buildLeafClusters(DBIDArrayIter it, int split, FiniteProgress progress) Prepare the leaf clusters by executing the first (size - 1 - split) merges.
Clustering<DendrogramModel>	extractClusters() Extract all clusters from the pi-lambda-representation.
protected abstract int	findSplit(DBIDArrayIter it) Find the splitting point in the ordered DBIDs list.
protected Cluster<DendrogramModel>	makeCluster(DBIDRef lead, double depth, DBIDs members) Make the cluster for the given object

Methods inherited from class java.lang.Object

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

Field Detail

ids

protected ArrayDBIDs ids

Unordered IDs

pi

protected DBIDDataStore pi

Parent pointer

lambda

protected DoubleDataStore lambda

Merge distance

pointerresult

protected PointerHierarchyRepresentationResult pointerresult

The hierarchical result to process.

cluster_map

protected WritableIntegerDataStore cluster_map

Map clusters to integer cluster numbers.

cluster_dbids

protected java.util.ArrayList<ModifiableDBIDs> cluster_dbids

Storage for cluster contents

clusterHeight

protected DoubleArray clusterHeight

Cluster distances

cluster_leads

protected ArrayModifiableDBIDs cluster_leads

Cluster lead objects

Constructor Detail

Instance

public Instance(PointerHierarchyRepresentationResult pointerresult)

Constructor.

Parameters:

pointerresult - Hierarchical result

Method Detail

extractClusters

public Clustering<DendrogramModel> extractClusters()

Extract all clusters from the pi-lambda-representation.

Returns:

Hierarchical clustering

buildLeafClusters

private void buildLeafClusters(DBIDArrayIter it,
                               int split,
                               FiniteProgress progress)

Prepare the leaf clusters by executing the first (size - 1 - split) merges.

Parameters:

it - Iterator

split - Splitting point

progress - Progress for logging (may be null)

buildFlat

private Clustering<DendrogramModel> buildFlat(DBIDArrayIter it,
                                              int split,
                                              FiniteProgress progress)

Build a flat clustering.

Parameters:

it - Ordered iterator

split - Splitting point

progress - Progress for logging (may be null)

Returns:

Clustering

buildHierarchical

private Clustering<DendrogramModel> buildHierarchical(DBIDArrayIter it,
                                                      int split,
                                                      FiniteProgress progress)

Build a hierarchical clustering.

Parameters:

it - Iterator

split - Splitting point

progress - Progress for logging (may be null)

Returns:

Clustering

findSplit

protected abstract int findSplit(DBIDArrayIter it)

Find the splitting point in the ordered DBIDs list.

Parameters:

it - Iterator on this list (reused)

Returns:

Splitting point

makeCluster

protected Cluster<DendrogramModel> makeCluster(DBIDRef lead,
                                               double depth,
                                               DBIDs members)

Make the cluster for the given object

Parameters:

lead - Leading object

depth - Linkage depth

members - Member objects

Returns:

Cluster