de.lmu.ifi.dbs.elki.math.statistics.dependence

Class SURFINGDependenceMeasure

java.lang.Object

de.lmu.ifi.dbs.elki.math.statistics.dependence.AbstractDependenceMeasure

de.lmu.ifi.dbs.elki.math.statistics.dependence.SURFINGDependenceMeasure

All Implemented Interfaces:

DependenceMeasure

@Priority(value=-100)
 @Reference(authors="Elke Achtert, Hans-Peter Kriegel, Erich Schubert, Arthur Zimek",title="Interactive Data Mining with 3D-Parallel-Coordinate-Trees",booktitle="Proc. 2013 ACM Int. Conf. on Management of Data (SIGMOD 2013)",url="https://doi.org/10.1145/2463676.2463696",bibkey="DBLP:conf/sigmod/AchtertKSZ13") @Reference(authors="Christian Baumgartner, Claudia Plant, Karin Kailing, Hans-Peter Kriegel, Peer Kr\u00f6ger",title="Subspace Selection for Clustering High-Dimensional Data",booktitle="Proc. IEEE International Conference on Data Mining (ICDM 2004)",url="https://doi.org/10.1109/ICDM.2004.10112",bibkey="DBLP:conf/icdm/BaumgartnerPKKK04")
public class SURFINGDependenceMeasure
extends AbstractDependenceMeasure

Compute the similarity of dimensions using the SURFING score. The parameter k for the k nearest neighbors is currently hard-coded to 10% of the set size.

Note that the complexity is roughly O(n n k), so this is a rather slow method, and with k at 10% of n, is actually cubic: O(0.1 * n²).

This version cannot use index support, as the API operates without database attachment. However, it should be possible to implement some trivial sorted-list indexes to get a reasonable speedup!

Reference:

Elke Achtert, Hans-Peter Kriegel, Erich Schubert, Arthur Zimek
Interactive Data Mining with 3D-Parallel-Coordinate-Trees
Proc. 2013 ACM Int. Conf. on Management of Data (SIGMOD 2013)

Based on:

Christian Baumgartner, Claudia Plant, Karin Kailing, Hans-Peter Kriegel, Peer Kröger
Subspace Selection for Clustering High-Dimensional Data
Proc. IEEE International Conference on Data Mining (ICDM 2004)

TODO: make the subspace distance function and k parameterizable.

TODO: results are not convincing, maybe try inserting points.

Since:

0.5.5

Author:

Robert Rödler, Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	SURFINGDependenceMeasure.Parameterizer Parameterization class.

Field Summary

Fields

Modifier and Type	Field and Description
static SURFINGDependenceMeasure	STATIC Static instance.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	SURFINGDependenceMeasure() Constructor.

Method Summary

Modifier and Type	Method and Description
<A,B> double	dependence(NumberArrayAdapter<?,A> adapter1, A data1, NumberArrayAdapter<?,B> adapter2, B data2) Measure the dependence of two variables.

Methods inherited from class de.lmu.ifi.dbs.elki.math.statistics.dependence.AbstractDependenceMeasure

clamp, computeNormalizedRanks, discretize, index, ranks, ranks, size, size, sortedIndex

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.math.statistics.dependence.DependenceMeasure

dependence, dependence, dependence

Field Detail

STATIC

public static final SURFINGDependenceMeasure STATIC

Static instance.

Constructor Detail

SURFINGDependenceMeasure

protected SURFINGDependenceMeasure()

Constructor. Use static instance instead!

Method Detail

dependence

public <A,B> double dependence(NumberArrayAdapter<?,A> adapter1,
                               A data1,
                               NumberArrayAdapter<?,B> adapter2,
                               B data2)

Description copied from interface: DependenceMeasure

Measure the dependence of two variables. This is the more flexible API, which allows using different internal data representations.

Type Parameters:

A - First array type

B - Second array type

Parameters:

adapter1 - First data adapter

data1 - First data set

adapter2 - Second data adapter

data2 - Second data set

Returns:

Dependence measure