jsat.clustering

Class TRIKMEDS

java.lang.Object

jsat.clustering.PAM

jsat.clustering.TRIKMEDS

All Implemented Interfaces:

java.io.Serializable, Clusterer, KClusterer

public class TRIKMEDS
extends PAM

This class implements the TRIKMEDS algorithm for PAM clustering. It returns the exact same result that would have been computed by PAM, but uses the triangle inequality to avoid unnecessary distance calculations. Expected runtime is O( n sqrt(n)), but still has worst case complexity O(n²). It also requires that the distance metric used be a valid distance metric.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
TRIKMEDS()
TRIKMEDS(DistanceMetric dm)
TRIKMEDS(DistanceMetric dm, java.util.Random rand)
TRIKMEDS(DistanceMetric dm, java.util.Random rand, SeedSelectionMethods.SeedSelection seedSelection)

Method Summary

Modifier and Type	Method and Description
static int	medoid(boolean parallel, java.util.Collection<java.lang.Integer> indecies, java.util.List<? extends Vec> X, DistanceMetric dm, java.util.List<java.lang.Double> accel) Computes the medoid of a sub-set of data
static int	medoid(boolean parallel, java.util.List<? extends Vec> X, DistanceMetric dm) Computes the medoid of the data
void	setDistanceMetric(DistanceMetric dm) Sets the distance metric used by this clustering algorithm

Methods inherited from class jsat.clustering.PAM

clone, cluster, cluster, cluster, getDistanceMetric, getMaxIterations, getMedoids, getSeedSelection, setMaxIterations, setSeedSelection, setStoreMedoids

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.clustering.KClusterer

cluster, cluster, cluster, cluster, cluster, cluster

Methods inherited from interface jsat.clustering.Clusterer

cluster, cluster, cluster, supportsWeightedData

Constructor Detail

TRIKMEDS

public TRIKMEDS(DistanceMetric dm,
                java.util.Random rand,
                SeedSelectionMethods.SeedSelection seedSelection)

TRIKMEDS

public TRIKMEDS(DistanceMetric dm,
                java.util.Random rand)

TRIKMEDS

public TRIKMEDS(DistanceMetric dm)

TRIKMEDS

public TRIKMEDS()

Method Detail

setDistanceMetric

public void setDistanceMetric(DistanceMetric dm)

Description copied from class: PAM

Sets the distance metric used by this clustering algorithm

Overrides:

setDistanceMetric in class PAM

Parameters:

dm - the distance metric to use

medoid

public static int medoid(boolean parallel,
                         java.util.List<? extends Vec> X,
                         DistanceMetric dm)

Computes the medoid of the data

Parameters:

parallel - whether or not the computation should be done using multiple cores

X - the list of all data

dm - the distance metric to get the medoid with respect to

Returns:

the index of the point in X that is the medoid

medoid

public static int medoid(boolean parallel,
                         java.util.Collection<java.lang.Integer> indecies,
                         java.util.List<? extends Vec> X,
                         DistanceMetric dm,
                         java.util.List<java.lang.Double> accel)

Computes the medoid of a sub-set of data

Parameters:

parallel - whether or not the computation should be done using multiple cores

indecies - the indexes of the points to get the medoid of

X - the list of all data

dm - the distance metric to get the medoid with respect to

accel - the acceleration cache for the distance metric

Returns:

the index value contained within indecies that is the medoid