Package elki.math.statistics.intrinsicdimensionality

Class RABIDEstimator

java.lang.Object

elki.math.statistics.intrinsicdimensionality.RABIDEstimator

All Implemented Interfaces:

DistanceBasedIntrinsicDimensionalityEstimator, IntrinsicDimensionalityEstimator<java.lang.Object>

Direct Known Subclasses:

ABIDEstimator

@Reference(authors="Erik Thordsen and Erich Schubert",
           title="ABID: Angle Based Intrinsic Dimensionality",
           booktitle="Proc. 13th Int. Conf. Similarity Search and Applications (SISAP\'2020)",
           url="https://doi.org/10.1007/978-3-030-60936-8_17",
           bibkey="DBLP:conf/sisap/ThordsenS20")
public class RABIDEstimator
extends java.lang.Object
implements DistanceBasedIntrinsicDimensionalityEstimator

Raw angle based intrinsic dimensionality (RABID) estimator.

Reference:

Erik Thordsen and Erich Schubert
ABID: Angle Based Intrinsic Dimensionality
Proc. 13th Int. Conf. Similarity Search and Applications (SISAP'2020)

Since:

0.8.0

Author:

Erik Thordsen

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	RABIDEstimator.Par	Parameterization class.

Field Summary

Fields

Modifier and Type	Field	Description
static RABIDEstimator	STATIC	Static instance.

Constructor Summary

Constructors

Constructor	Description
RABIDEstimator()

Method Summary

Modifier and Type	Method	Description
protected double	computeABID(DistanceQuery<?> distq, KNNList knn, boolean bias)	Estimate intrinsic dimensionality (both variants).
private double	cos2(double sideA2, double sideB2, double oppositeSide2)	Squared cosine from squared triangle side lengths
<A> double	estimate(A data, NumberArrayAdapter<?,? super A> adapter, int size)	Estimate from a distance list.
double	estimate(KNNSearcher<DBIDRef> knnq, DistanceQuery<?> distq, DBIDRef cur, int k)	Estimate from a Reference Point, a KNNSearcher and the neighborhood size k.

Methods inherited from class java.lang.Object

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

Methods inherited from interface elki.math.statistics.intrinsicdimensionality.DistanceBasedIntrinsicDimensionalityEstimator

estimate, estimate, estimate, estimate

Field Detail

STATIC

public static final RABIDEstimator STATIC

Static instance.

Constructor Detail

RABIDEstimator

public RABIDEstimator()

Method Detail

estimate

public <A> double estimate(A data,
                           NumberArrayAdapter<?,? super A> adapter,
                           int size)

Description copied from interface: DistanceBasedIntrinsicDimensionalityEstimator

Estimate from a distance list.

Specified by:

estimate in interface DistanceBasedIntrinsicDimensionalityEstimator

Type Parameters:

A - array type

Parameters:

data - Data

adapter - Array adapter

size - Length

Returns:

Estimated intrinsic dimensionality

cos2

private final double cos2(double sideA2,
                          double sideB2,
                          double oppositeSide2)

Squared cosine from squared triangle side lengths

Parameters:

sideA2 - squared first side

sideB2 - squared second side

oppositeSide2 - squared opposite side

Returns:

Squared cosine

estimate

public double estimate(KNNSearcher<DBIDRef> knnq,
                       DistanceQuery<?> distq,
                       DBIDRef cur,
                       int k)

Description copied from interface: IntrinsicDimensionalityEstimator

Estimate from a Reference Point, a KNNSearcher and the neighborhood size k.

Specified by:

estimate in interface DistanceBasedIntrinsicDimensionalityEstimator

Specified by:

estimate in interface IntrinsicDimensionalityEstimator<java.lang.Object>

Parameters:

knnq - KNNSearcher

distq - Distance query for additional distances

cur - reference point

k - neighborhood size

Returns:

Estimated intrinsic dimensionality

computeABID

protected double computeABID(DistanceQuery<?> distq,
                             KNNList knn,
                             boolean bias)

Estimate intrinsic dimensionality (both variants).

Parameters:

distq - Distance query

knn - k nearest neighbors (usually +1, as the query point is included)

bias - true to use ABID, false for RABID

Returns:

intrinsic dimensionality