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

Class SLINKHDBSCANLinearMemory<O>

java.lang.Object

de.lmu.ifi.dbs.elki.algorithm.AbstractAlgorithm<R>

de.lmu.ifi.dbs.elki.algorithm.AbstractDistanceBasedAlgorithm<O,R>

de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.AbstractHDBSCAN<O,PointerDensityHierarchyRepresentationResult>

de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.SLINKHDBSCANLinearMemory<O>

All Implemented Interfaces:

Algorithm, HierarchicalClusteringAlgorithm, DistanceBasedAlgorithm<O>

@Title(value="HDBSCAN: Hierarchical Density-Based Spatial Clustering of Applications with Noise")
@Description(value="Density-Based Clustering Based on Hierarchical Density Estimates")
@Reference(authors="R. J. G. B. Campello, D. Moulavi, and J. Sander",
           title="Density-Based Clustering Based on Hierarchical Density Estimates",
           booktitle="Pacific-Asia Conference on Advances in Knowledge Discovery and Data Mining, PAKDD",
           url="http://dx.doi.org/10.1007/978-3-642-37456-2_14")
public class SLINKHDBSCANLinearMemory<O>
extends AbstractHDBSCAN<O,PointerDensityHierarchyRepresentationResult>
implements HierarchicalClusteringAlgorithm

Linear memory implementation of HDBSCAN clustering based on SLINK. By not building a distance matrix, we can reduce memory usage to linear memory only; but at the cost of roughly double the runtime (unless using indexes) as we first need to compute all kNN distances (for core sizes), then recompute distances when building the spanning tree. This version uses the SLINK algorithm to directly produce the pointer representation expected by the extraction methods. The SLINK algorithm is closely related to Prim's minimum spanning tree, but produces the more compact pointer representation instead of an edges list. This implementation does not include the cluster extraction discussed as Step 4. This functionality should however already be provided by HDBSCANHierarchyExtraction . For this reason, we also do not include self-edges. Reference:

R. J. G. B. Campello, D. Moulavi, and J. Sander
Density-Based Clustering Based on Hierarchical Density Estimates
Pacific-Asia Conference on Advances in Knowledge Discovery and Data Mining, PAKDD

Author:

Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	SLINKHDBSCANLinearMemory.Parameterizer<O> Parameterization class

Nested classes/interfaces inherited from class de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.AbstractHDBSCAN

AbstractHDBSCAN.HDBSCANAdapter, AbstractHDBSCAN.HeapMSTCollector

Field Summary

Fields

Modifier and Type	Field and Description
private static Logging	LOG Class logger.

Fields inherited from class de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.AbstractHDBSCAN

minPts

Fields inherited from interface de.lmu.ifi.dbs.elki.algorithm.DistanceBasedAlgorithm

DISTANCE_FUNCTION_ID

Constructor Summary

Constructors

Constructor and Description
SLINKHDBSCANLinearMemory(DistanceFunction<? super O> distanceFunction, int minPts) Constructor.

Method Summary

Methods

Modifier and Type	Method and Description
TypeInformation[]	getInputTypeRestriction() Get the input type restriction used for negotiating the data query.
protected Logging	getLogger() Get the (STATIC) logger for this class.
PointerDensityHierarchyRepresentationResult	run(Database db, Relation<O> relation) Run the algorithm
private void	step1(DBIDRef id, WritableDBIDDataStore pi, WritableDoubleDataStore lambda) First step: Initialize P(id) = id, L(id) = infinity.
private void	step2(DBIDRef id, DBIDs processedIDs, DistanceQuery<? super O> distQuery, DoubleDataStore coredists, WritableDoubleDataStore m) Second step: Determine the pairwise distances from all objects in the pointer representation to the new object with the specified id.
private void	step3(DBIDRef id, WritableDBIDDataStore pi, WritableDoubleDataStore lambda, DBIDs processedIDs, WritableDoubleDataStore m) Third step: Determine the values for P and L
private void	step4(DBIDRef id, WritableDBIDDataStore pi, WritableDoubleDataStore lambda, DBIDs processedIDs) Fourth step: Actualize the clusters if necessary

Methods inherited from class de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.AbstractHDBSCAN

computeCoreDists, convertToPointerRepresentation

Methods inherited from class de.lmu.ifi.dbs.elki.algorithm.AbstractDistanceBasedAlgorithm

getDistanceFunction

Methods inherited from class de.lmu.ifi.dbs.elki.algorithm.AbstractAlgorithm

makeParameterDistanceFunction, run

Methods inherited from class java.lang.Object

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

Methods inherited from interface de.lmu.ifi.dbs.elki.algorithm.clustering.hierarchical.HierarchicalClusteringAlgorithm

run

Field Detail

LOG

private static final Logging LOG

Class logger.

Constructor Detail

SLINKHDBSCANLinearMemory

public SLINKHDBSCANLinearMemory(DistanceFunction<? super O> distanceFunction,
                        int minPts)

Constructor.

Parameters:

distanceFunction - Distance function

minPts - Minimum number of points for density

Method Detail

run

public PointerDensityHierarchyRepresentationResult run(Database db,
                                              Relation<O> relation)

Run the algorithm

Parameters:

db - Database

relation - Relation

Returns:

Clustering hierarchy

step1

private void step1(DBIDRef id,
         WritableDBIDDataStore pi,
         WritableDoubleDataStore lambda)

First step: Initialize P(id) = id, L(id) = infinity.

Parameters:

id - the id of the object to be inserted into the pointer representation

pi - Pi data store

lambda - Lambda data store

step2

private void step2(DBIDRef id,
         DBIDs processedIDs,
         DistanceQuery<? super O> distQuery,
         DoubleDataStore coredists,
         WritableDoubleDataStore m)

Second step: Determine the pairwise distances from all objects in the pointer representation to the new object with the specified id.

Parameters:

id - the id of the object to be inserted into the pointer representation

processedIDs - the already processed ids

distQuery - Distance query

m - Data store

step3

private void step3(DBIDRef id,
         WritableDBIDDataStore pi,
         WritableDoubleDataStore lambda,
         DBIDs processedIDs,
         WritableDoubleDataStore m)

Third step: Determine the values for P and L

Parameters:

id - the id of the object to be inserted into the pointer representation

pi - Pi data store

lambda - Lambda data store

processedIDs - the already processed ids

m - Data store

step4

private void step4(DBIDRef id,
         WritableDBIDDataStore pi,
         WritableDoubleDataStore lambda,
         DBIDs processedIDs)

Fourth step: Actualize the clusters if necessary

Parameters:

id - the id of the current object

pi - Pi data store

lambda - Lambda data store

processedIDs - the already processed ids

getInputTypeRestriction

public TypeInformation[] getInputTypeRestriction()

Description copied from class: AbstractAlgorithm

Get the input type restriction used for negotiating the data query.

Specified by:

getInputTypeRestriction in interface Algorithm

Overrides:

getInputTypeRestriction in class AbstractHDBSCAN<O,PointerDensityHierarchyRepresentationResult>

Returns:

Type restriction

getLogger

protected Logging getLogger()

Description copied from class: AbstractAlgorithm

Get the (STATIC) logger for this class.

Specified by:

getLogger in class AbstractAlgorithm<PointerDensityHierarchyRepresentationResult>

Returns:

the static logger