de.lmu.ifi.dbs.elki.database.query.knn

Class LinearScanRawDoubleDistanceKNNQuery<O>

java.lang.Object

de.lmu.ifi.dbs.elki.database.query.AbstractDataBasedQuery<O>

de.lmu.ifi.dbs.elki.database.query.knn.AbstractDistanceKNNQuery<O,D>

de.lmu.ifi.dbs.elki.database.query.knn.LinearScanKNNQuery<O,D>

de.lmu.ifi.dbs.elki.database.query.knn.LinearScanPrimitiveDistanceKNNQuery<O,DoubleDistance>

de.lmu.ifi.dbs.elki.database.query.knn.LinearScanRawDoubleDistanceKNNQuery<O>

Type Parameters:

O - Object type

All Implemented Interfaces:

DatabaseQuery, KNNQuery<O,DoubleDistance>, LinearScanQuery

public class LinearScanRawDoubleDistanceKNNQuery<O>
extends LinearScanPrimitiveDistanceKNNQuery<O,DoubleDistance>

Optimized linear scan query for PrimitiveDoubleDistanceFunctions.

Field Summary

Fields

Modifier and Type	Field and Description
(package private) PrimitiveDoubleDistanceFunction<O>	rawdist Raw distance function.

Fields inherited from class de.lmu.ifi.dbs.elki.database.query.knn.AbstractDistanceKNNQuery

distanceQuery

Fields inherited from class de.lmu.ifi.dbs.elki.database.query.AbstractDataBasedQuery

relation

Fields inherited from interface de.lmu.ifi.dbs.elki.database.query.DatabaseQuery

HINT_BULK, HINT_EXACT, HINT_HEAVY_USE, HINT_NO_CACHE, HINT_OPTIMIZED_ONLY, HINT_SINGLE

Constructor Summary

Constructors

Constructor and Description
LinearScanRawDoubleDistanceKNNQuery(PrimitiveDistanceQuery<O,DoubleDistance> distanceQuery) Constructor.

Method Summary

Methods

Modifier and Type	Method and Description
KNNResult<DoubleDistance>	getKNNForDBID(DBIDRef id, int k) Get the k nearest neighbors for a particular id.
KNNResult<DoubleDistance>	getKNNForObject(O obj, int k) Get the k nearest neighbors for a particular id.
(package private) KNNResult<DoubleDistance>	getKNNForObjectBenchmarked(O obj, int k) It does not make sense, but this version is faster in our larger benchmarks.
(package private) KNNResult<DoubleDistance>	getKNNForObjectClean(O obj, int k) This is the cleaner, supposedly faster implementation.

Methods inherited from class de.lmu.ifi.dbs.elki.database.query.knn.LinearScanPrimitiveDistanceKNNQuery

getKNNForBulkDBIDs, linearScanBatchKNN

Methods inherited from class de.lmu.ifi.dbs.elki.database.query.AbstractDataBasedQuery

getRelation

Methods inherited from class java.lang.Object

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

Field Detail

rawdist

PrimitiveDoubleDistanceFunction<O> rawdist

Raw distance function.

Constructor Detail

LinearScanRawDoubleDistanceKNNQuery

public LinearScanRawDoubleDistanceKNNQuery(PrimitiveDistanceQuery<O,DoubleDistance> distanceQuery)

Constructor.

Parameters:

distanceQuery - Distance function to use

Method Detail

getKNNForDBID

public KNNResult<DoubleDistance> getKNNForDBID(DBIDRef id,
                                      int k)

Description copied from interface: KNNQuery

Get the k nearest neighbors for a particular id.

Specified by:

getKNNForDBID in interface KNNQuery<O,DoubleDistance>

Overrides:

getKNNForDBID in class LinearScanPrimitiveDistanceKNNQuery<O,DoubleDistance>

Parameters:

id - query object ID

k - Number of neighbors requested

Returns:

neighbors

getKNNForObject

public KNNResult<DoubleDistance> getKNNForObject(O obj,
                                        int k)

Description copied from interface: KNNQuery

Get the k nearest neighbors for a particular id.

Specified by:

getKNNForObject in interface KNNQuery<O,DoubleDistance>

Overrides:

getKNNForObject in class LinearScanKNNQuery<O,DoubleDistance>

Parameters:

obj - Query object

k - Number of neighbors requested

Returns:

neighbors

getKNNForObjectClean

KNNResult<DoubleDistance> getKNNForObjectClean(O obj,
                                             int k)

This is the cleaner, supposedly faster implementation.

Parameters:

obj - Query object

k - Desired number of neighbors

Returns:

kNN result

getKNNForObjectBenchmarked

KNNResult<DoubleDistance> getKNNForObjectBenchmarked(O obj,
                                                   int k)

It does not make sense, but this version is faster in our larger benchmarks. Apparently, some JIT optimization kicks in better.

Parameters:

obj - Query object

k - Desired number of neighbors

Returns:

kNN result