Package elki.clustering.hierarchical

Class HACAM.Instance

java.lang.Object

elki.clustering.hierarchical.AGNES.Instance

elki.clustering.hierarchical.Anderberg.Instance

elki.clustering.hierarchical.HACAM.Instance

Enclosing class:

HACAM<O>

public static class HACAM.Instance
extends Anderberg.Instance

Main worker instance of AGNES.

Author:

Erich Schubert

Field Summary

Fields

Modifier and Type	Field	Description
protected it.unimi.dsi.fastutil.ints.Int2ObjectOpenHashMap<ModifiableDBIDs>	clusters	Cluster to members map
protected DistanceQuery<?>	dq	Distance query
protected DBIDArrayIter	ix	Iterators into the object ids.
protected DBIDArrayIter	iy	Iterators into the object ids.
protected DBIDArrayMIter	prots	Iterator into the prototypes
protected double[]	tds	Total deviations (for minimum sum increase only)
protected HACAM.Variant	variant	Linkage variant to use

Fields inherited from class elki.clustering.hierarchical.Anderberg.Instance

bestd, besti

Fields inherited from class elki.clustering.hierarchical.AGNES.Instance

builder, end, linkage, mat

Constructor Summary

Constructors

Constructor	Description
Instance(HACAM.Variant variant)	Constructor.

Method Summary

Modifier and Type	Method	Description
private static double	distanceSum(DistanceQuery<?> dq, DBIDIter i, DBIDs cy, double distsum, double minDistSum)	Find the maximum distance of one object to a set.
protected int	findMerge()	Perform the next merge step.
private static double	findPrototype(DistanceQuery<?> dq, DBIDs cx, DBIDs cy, DBIDVar prototype, double minDistSum)	Find the prototypes.
private static double	findPrototypeSingleton(DistanceQuery<?> dq, DBIDs cx, DBIDRef cy, DBIDVar prototype)	Find the prototypes.
protected void	merge(int x, int y)	Execute the cluster merge.
ClusterMergeHistory	run(ClusterDistanceMatrix mat, ClusterMergeHistoryBuilder builder)	Run the main algorithm.
ClusterPrototypeMergeHistory	run(ArrayDBIDs ids, ClusterDistanceMatrix mat, ClusterMergeHistoryBuilder builder, DistanceQuery<?> dq, DBIDArrayMIter prots)	Run HACAM linkage
protected void	updateEntry(int x, int y)	Update entry at x,y for distance matrix distances
private void	updateMatrices(int x, int y)	Update the entries of the matrices that contain a distance to y, the newly merged cluster.

Methods inherited from class elki.clustering.hierarchical.Anderberg.Instance

findBest, initializeNNCache, merge, updateCache, updateMatrix

Methods inherited from class elki.clustering.hierarchical.AGNES.Instance

shrinkActiveSet

Methods inherited from class java.lang.Object

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

Field Detail

variant

protected HACAM.Variant variant

Linkage variant to use

clusters

protected it.unimi.dsi.fastutil.ints.Int2ObjectOpenHashMap<ModifiableDBIDs> clusters

Cluster to members map

tds

protected double[] tds

Total deviations (for minimum sum increase only)

dq

protected DistanceQuery<?> dq

Distance query

prots

protected DBIDArrayMIter prots

Iterator into the prototypes

ix

protected DBIDArrayIter ix

Iterators into the object ids.

iy

protected DBIDArrayIter iy

Iterators into the object ids.

Constructor Detail

Instance

public Instance(HACAM.Variant variant)

Constructor.

Parameters:

variant - HACAM variant to use

Method Detail

run

public ClusterMergeHistory run(ClusterDistanceMatrix mat,
                               ClusterMergeHistoryBuilder builder)

Description copied from class: AGNES.Instance

Run the main algorithm.

Overrides:

run in class Anderberg.Instance

Parameters:

mat - Distance matrix

builder - Result builder

Returns:

Cluster history

run

public ClusterPrototypeMergeHistory run(ArrayDBIDs ids,
                                        ClusterDistanceMatrix mat,
                                        ClusterMergeHistoryBuilder builder,
                                        DistanceQuery<?> dq,
                                        DBIDArrayMIter prots)

Run HACAM linkage

Parameters:

ids - Object ids

mat - Distance matrix

builder - Result builder

dq - Distance query

prots - Iterator into prototypes

Returns:

Cluster merge history

findMerge

protected int findMerge()

Description copied from class: Anderberg.Instance

Perform the next merge step.

Due to the cache, this is now O(n) each time, instead of O(n*n).

Overrides:

findMerge in class Anderberg.Instance

Returns:

x, for shrinking the working set.

merge

protected void merge(int x,
                     int y)

Execute the cluster merge.

Parameters:

x - first cluster to merge, with x > y

y - second cluster to merge, with y < x

updateMatrices

private void updateMatrices(int x,
                            int y)

Update the entries of the matrices that contain a distance to y, the newly merged cluster.

Parameters:

x - first cluster to merge, with x > y

y - second cluster to merge, with y < x

updateEntry

protected void updateEntry(int x,
                           int y)

Update entry at x,y for distance matrix distances

Parameters:

x - index of cluster, x > y

y - index of cluster, y < x

findPrototype

private static double findPrototype(DistanceQuery<?> dq,
                                    DBIDs cx,
                                    DBIDs cy,
                                    DBIDVar prototype,
                                    double minDistSum)

Find the prototypes.

Parameters:

dq - Distance query

cx - First set

cy - Second set

prototype - Prototype output variable

minDistSum - Previously best distance.

Returns:

New best distance

findPrototypeSingleton

private static double findPrototypeSingleton(DistanceQuery<?> dq,
                                             DBIDs cx,
                                             DBIDRef cy,
                                             DBIDVar prototype)

Find the prototypes.

Parameters:

dq - Distance query

cx - First set

cy - Singleton object

prototype - Prototype output variable

Returns:

New best distance

distanceSum

private static double distanceSum(DistanceQuery<?> dq,
                                  DBIDIter i,
                                  DBIDs cy,
                                  double distsum,
                                  double minDistSum)

Find the maximum distance of one object to a set.

Parameters:

dq - Distance query

i - Current object

cy - Set of candidates

distsum - Current sum

minDistSum - Early stopping threshold

Returns:

Distance sum