Package elki.algorithm.statistics

Class EvaluateRetrievalPerformance<O>

java.lang.Object

elki.algorithm.statistics.EvaluateRetrievalPerformance<O>

Type Parameters:

O - Object type

All Implemented Interfaces:

Algorithm

public class EvaluateRetrievalPerformance<O>
extends java.lang.Object
implements Algorithm

Evaluate a distance functions performance by computing the mean average precision, ROC, and NN classification performance when ranking the objects by distance.

Since:

0.7.0

Author:

Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	EvaluateRetrievalPerformance.KNNEvaluator	Evaluate kNN retrieval performance.
static class	EvaluateRetrievalPerformance.RetrievalPerformanceResult	Result object for MAP scores.

Nested classes/interfaces inherited from interface elki.Algorithm

Algorithm.Utils

Field Summary

Fields

Modifier and Type	Field	Description
protected Distance<? super O>	distance	Distance function used.
protected boolean	includeSelf	Include query object in evaluation.
private static Logging	LOG	The logger for this class.
protected int	maxk	K nearest neighbors to use for classification evaluation.
private java.lang.String	PREFIX	Prefix for statistics.
protected RandomFactory	random	Random sampling seed.
protected double	sampling	Relative number of object to use in sampling.

Constructor Summary

Constructors

Constructor	Description
EvaluateRetrievalPerformance(Distance<? super O> distance, double sampling, RandomFactory random, boolean includeSelf, int maxk)	Constructor.

Method Summary

Modifier and Type	Method	Description
private void	computeDistances(ModifiableDoubleDBIDList nlist, DBIDIter query, DistanceQuery<O> distQuery, Relation<O> relation)	Compute the distances to the neighbor objects.
private void	findMatches(ModifiableDBIDs posn, Relation<?> lrelation, java.lang.Object label)	Find all matching objects.
TypeInformation[]	getInputTypeRestriction()	Get the input type restriction used for negotiating the data query.
protected static boolean	match(java.lang.Object ref, java.lang.Object test)	Test whether two relation agree.
EvaluateRetrievalPerformance.RetrievalPerformanceResult	run(Relation<O> relation, Relation<?> lrelation)	Run the algorithm

Methods inherited from class java.lang.Object

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

Methods inherited from interface elki.Algorithm

autorun

Field Detail

LOG

private static final Logging LOG

The logger for this class.

PREFIX

private final java.lang.String PREFIX

Prefix for statistics.

distance

protected Distance<? super O> distance

Distance function used.

sampling

protected double sampling

Relative number of object to use in sampling.

random

protected RandomFactory random

Random sampling seed.

includeSelf

protected boolean includeSelf

Include query object in evaluation.

maxk

protected int maxk

K nearest neighbors to use for classification evaluation.

Constructor Detail

EvaluateRetrievalPerformance

public EvaluateRetrievalPerformance(Distance<? super O> distance,
                                    double sampling,
                                    RandomFactory random,
                                    boolean includeSelf,
                                    int maxk)

Constructor.

Parameters:

distance - Distance function

sampling - Sampling rate

random - Random sampling generator

includeSelf - Include query object in evaluation

maxk - Maximum k for kNN evaluation

Method Detail

getInputTypeRestriction

public TypeInformation[] getInputTypeRestriction()

Description copied from interface: Algorithm

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

Specified by:

getInputTypeRestriction in interface Algorithm

Returns:

Type restriction

run

public EvaluateRetrievalPerformance.RetrievalPerformanceResult run(Relation<O> relation,
                                                                   Relation<?> lrelation)

Run the algorithm

Parameters:

relation - Relation for distance computations

lrelation - Relation for class label comparison

Returns:

Vectors containing mean and standard deviation.

match

protected static boolean match(java.lang.Object ref,
                               java.lang.Object test)

Test whether two relation agree.

Parameters:

ref - Reference object

test - Test object

Returns:

true if the objects match

findMatches

private void findMatches(ModifiableDBIDs posn,
                         Relation<?> lrelation,
                         java.lang.Object label)

Find all matching objects.

Parameters:

posn - Output set.

lrelation - Label relation

label - Query object label

computeDistances

private void computeDistances(ModifiableDoubleDBIDList nlist,
                              DBIDIter query,
                              DistanceQuery<O> distQuery,
                              Relation<O> relation)

Compute the distances to the neighbor objects.

Parameters:

nlist - Neighbor list (output)

query - Query object

distQuery - Distance function

relation - Data relation