Package elki.clustering.kmeans

Class AbstractKMeans.Par<V extends NumberVector>

java.lang.Object

elki.clustering.kmeans.AbstractKMeans.Par<V>

All Implemented Interfaces:

Parameterizer

Direct Known Subclasses:

BetulaLloydKMeans.Par, CompareMeans.Par, GMeans.Par, HamerlyKMeans.Par, HartiganWongKMeans.Parameterizer, KDTreePruningKMeans.Par, KMeansMinusMinus.Par, KMediansLloyd.Par, LloydKMeans.Par, MacQueenKMeans.Par, SimplifiedElkanKMeans.Par, SingleAssignmentKMeans.Par, SortMeans.Par, SphericalKMeans.Par, YinYangKMeans.Par

Enclosing class:

AbstractKMeans<V extends NumberVector,M extends Model>

public abstract static class AbstractKMeans.Par<V extends NumberVector>
extends java.lang.Object
implements Parameterizer

Parameterization class.

Author:

Erich Schubert

Field Summary

Fields

Modifier and Type	Field	Description
protected NumberVectorDistance<? super V>	distance	The distance function to use.
protected KMeansInitialization	initializer	Initialization method.
protected int	k	k Parameter.
protected int	maxiter	Maximum number of iterations.
protected boolean	varstat	Compute the final variance statistic (not used by all).

Constructor Summary

Constructors

Constructor	Description
Par()

Method Summary

Modifier and Type	Method	Description
void	configure(Parameterization config)	Configure the class.
protected void	getParameterDistance(Parameterization config)	Get the distance function parameter.
protected void	getParameterInitialization(Parameterization config)	Get the initialization method parameter.
protected void	getParameterK(Parameterization config)	Get the k parameter.
protected void	getParameterMaxIter(Parameterization config)	Get the max iterations parameter.
protected void	getParameterVarstat(Parameterization config)	Get the variance statistics parameter.
abstract AbstractKMeans<V,?>	make()	Make an instance after successful configuration.
protected boolean	needsMetric()	Users could use other non-metric distances at their own risk; but some k-means variants make explicit use of the triangle inequality, we emit extra warnings then.

Methods inherited from class java.lang.Object

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

Field Detail

k

protected int k

k Parameter.

maxiter

protected int maxiter

Maximum number of iterations.

initializer

protected KMeansInitialization initializer

Initialization method.

varstat

protected boolean varstat

Compute the final variance statistic (not used by all).

distance

protected NumberVectorDistance<? super V extends NumberVector> distance

The distance function to use.

Constructor Detail

Par

public Par()

Method Detail

configure

public void configure(Parameterization config)

Description copied from interface: Parameterizer

Configure the class.

Note: the status is collected by the parameterization object, so that multiple errors may arise and be reported in one run.

Specified by:

configure in interface Parameterizer

Parameters:

config - Parameterization

getParameterK

protected void getParameterK(Parameterization config)

Get the k parameter.

Parameters:

config - Parameterization

getParameterDistance

protected void getParameterDistance(Parameterization config)

Get the distance function parameter.

Parameters:

config - Parameterization

needsMetric

protected boolean needsMetric()

Users could use other non-metric distances at their own risk; but some k-means variants make explicit use of the triangle inequality, we emit extra warnings then.

Returns:

true if the algorithm uses triangle inequality

getParameterInitialization

protected void getParameterInitialization(Parameterization config)

Get the initialization method parameter.

Parameters:

config - Parameterization

getParameterMaxIter

protected void getParameterMaxIter(Parameterization config)

Get the max iterations parameter.

Parameters:

config - Parameterization

getParameterVarstat

protected void getParameterVarstat(Parameterization config)

Get the variance statistics parameter.

Parameters:

config - Parameterization

make

public abstract AbstractKMeans<V,?> make()

Description copied from interface: Parameterizer

Make an instance after successful configuration.

Note: your class should return the exact type, only this very broad interface should use Object as return type.

Specified by:

make in interface Parameterizer

Returns:

a new instance