de.lmu.ifi.dbs.elki.datasource.filter.transform

Class ClassicMultidimensionalScalingTransform<O>

java.lang.Object

de.lmu.ifi.dbs.elki.datasource.filter.transform.ClassicMultidimensionalScalingTransform<O>

Type Parameters:

O - Data type

All Implemented Interfaces:

ObjectFilter

@Alias(value="mds")
public class ClassicMultidimensionalScalingTransform<O>
extends Object
implements ObjectFilter

Rescale the data set using multidimensional scaling, MDS. Note: the current implementation is rather expensive, both memory- and runtime wise. Don't use for large data sets! Instead, have a look at FastMultidimensionalScalingTransform which uses power iterations instead.

Since:

0.6.0

Author:

Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ClassicMultidimensionalScalingTransform.Parameterizer<O extends NumberVector> Parameterization class.

Field Summary

Fields

Modifier and Type	Field and Description
(package private) PrimitiveDistanceFunction<? super O>	dist Distance function to use.
private static Logging	LOG Class logger.
(package private) int	tdim Target dimensionality

Constructor Summary

Constructors

Constructor and Description
ClassicMultidimensionalScalingTransform(int tdim, PrimitiveDistanceFunction<? super O> dist) Constructor.

Method Summary

Methods

Modifier and Type	Method and Description
protected double[][]	computeDistanceMatrix(List<O> castColumn, int size)
static void	doubleCenterSymmetric(double[][] m) Double-center the given matrix (only upper triangle is used).
MultipleObjectsBundle	filter(MultipleObjectsBundle objects) Filter a set of object packages.

Methods inherited from class java.lang.Object

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

Field Detail

LOG

private static final Logging LOG

Class logger.

dist

PrimitiveDistanceFunction<? super O> dist

Distance function to use.

tdim

int tdim

Target dimensionality

Constructor Detail

ClassicMultidimensionalScalingTransform

public ClassicMultidimensionalScalingTransform(int tdim,
                                       PrimitiveDistanceFunction<? super O> dist)

Constructor.

Parameters:

tdim - Target dimensionality.

dist - Distance function to use.

Method Detail

filter

public MultipleObjectsBundle filter(MultipleObjectsBundle objects)

Description copied from interface: ObjectFilter

Filter a set of object packages.

Specified by:

filter in interface ObjectFilter

Parameters:

objects - Object to filter

Returns:

Filtered objects

computeDistanceMatrix

protected double[][] computeDistanceMatrix(List<O> castColumn,
                               int size)

doubleCenterSymmetric

public static void doubleCenterSymmetric(double[][] m)

Double-center the given matrix (only upper triangle is used). For improved numerical precision, we perform incremental updates to the mean values, instead of computing a large sum and then performing division.

Parameters:

m - Matrix to double-center.