de.lmu.ifi.dbs.elki.preprocessing
Class LocalPCAPreprocessor<V extends NumberVector<V,?>>

java.lang.Object
  de.lmu.ifi.dbs.elki.logging.AbstractLoggable
      de.lmu.ifi.dbs.elki.preprocessing.LocalPCAPreprocessor<V>

Type Parameters:

V - the type of NumberVector handled by this Preprocessor

All Implemented Interfaces:

Preprocessor<V>

Direct Known Subclasses:

KnnQueryBasedLocalPCAPreprocessor, RangeQueryBasedLocalPCAPreprocessor

@Title(value="Local PCA Preprocessor")
@Description(value="Materializes the local PCA and the locally weighted matrix of objects of a database.")
public abstract class LocalPCAPreprocessor<V extends NumberVector<V,?>>
extends AbstractLoggable
implements Preprocessor<V>

Abstract superclass for preprocessors performing for each object of a certain database a filtered PCA based on the local neighborhood of the object. The result of the PCA is assigned to the database with association id AssociationID.LOCAL_PCA. Additionally, a copy of the PCAs similarity matrix is assigned with association id AssociationID.LOCALLY_WEIGHTED_MATRIX.

Author:

Elke Achtert

Field Summary

static Class<?>	DEFAULT_PCA_DISTANCE_FUNCTION The default distance function for the PCA.
private PCAFilteredRunner<V,DoubleDistance>	pca PCA utility object.
static OptionID	PCA_DISTANCE_ID OptionID for PCA_DISTANCE_PARAM
protected ObjectParameter<DistanceFunction<V,DoubleDistance>>	PCA_DISTANCE_PARAM Parameter to specify the distance function used for running PCA.
protected DistanceFunction<V,DoubleDistance>	pcaDistanceFunction Holds the instance of the distance function specified by PCA_DISTANCE_PARAM.

Fields inherited from class de.lmu.ifi.dbs.elki.logging.AbstractLoggable

debug, logger

Constructor Summary

LocalPCAPreprocessor(Parameterization config)
Constructor, adhering to Parameterizable

Method Summary

protected abstract List<DistanceResultPair<DoubleDistance>>	objectsForPCA(Integer id, Database<V> database) Returns the objects to be considered within the PCA for the specified query object.
void	run(Database<V> database, boolean verbose, boolean time) Performs for each object of the specified database a filtered PCA based on the local neighborhood of the object.

Methods inherited from class de.lmu.ifi.dbs.elki.logging.AbstractLoggable

debugFine, debugFiner, debugFinest, exception, progress, verbose, warning

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_PCA_DISTANCE_FUNCTION

public static final Class<?> DEFAULT_PCA_DISTANCE_FUNCTION

The default distance function for the PCA.

PCA_DISTANCE_ID

public static final OptionID PCA_DISTANCE_ID

OptionID for PCA_DISTANCE_PARAM

PCA_DISTANCE_PARAM

protected final ObjectParameter<DistanceFunction<V extends NumberVector<V,?>,DoubleDistance>> PCA_DISTANCE_PARAM

Parameter to specify the distance function used for running PCA. Key: -localpca.distancefunction

pcaDistanceFunction

protected DistanceFunction<V extends NumberVector<V,?>,DoubleDistance> pcaDistanceFunction

Holds the instance of the distance function specified by PCA_DISTANCE_PARAM.

pca

private PCAFilteredRunner<V extends NumberVector<V,?>,DoubleDistance> pca

PCA utility object.

Constructor Detail

LocalPCAPreprocessor

public LocalPCAPreprocessor(Parameterization config)

Constructor, adhering to Parameterizable

Parameters:

config - Parameterization

Method Detail

run

public void run(Database<V> database,
                boolean verbose,
                boolean time)

Performs for each object of the specified database a filtered PCA based on the local neighborhood of the object. The result of the PCA is assigned to the database with association id AssociationID.LOCAL_PCA. Additionally, a copy of the PCAs similarity matrix is assigned with association id AssociationID.LOCALLY_WEIGHTED_MATRIX.

Specified by:

run in interface Preprocessor<V extends NumberVector<V,?>>

Parameters:

database - the database for which the preprocessing is performed

verbose - flag to allow verbose messages while performing the algorithm

time - flag to request output of performance time

objectsForPCA

protected abstract List<DistanceResultPair<DoubleDistance>> objectsForPCA(Integer id,
                                                                          Database<V> database)

Returns the objects to be considered within the PCA for the specified query object.

Parameters:

id - the id of the query object for which a PCA should be performed

database - the database holding the objects

Returns:

the list of the objects (i.e. the ids and the distances to the query object) to be considered within the PCA