weka.clusterers

Class CentroidSketch

java.lang.Object

weka.clusterers.CentroidSketch

All Implemented Interfaces:

java.io.Serializable, weka.core.TechnicalInformationHandler

public class CentroidSketch
extends java.lang.Object
implements weka.core.TechnicalInformationHandler, java.io.Serializable

Class for managing a sketch of centres for k-means, along with a weighted reservoir sample that is used over iterations to update the sketch. Used in the implementation of the k-means|| initialization method. For more information, see

Bahman Bahmani, Benjamin Moseley, Andrea Vattani, Ravi Kumar, Sergei Vassilvitskii (2012). Scalable k-means++. Proceedings of the VLDB Endowment.:622-633.

BibTeX:

 @article{Bahmani2012,
    author = {Bahman Bahmani and Benjamin Moseley and Andrea Vattani and Ravi Kumar and Sergei Vassilvitskii},
    journal = {Proceedings of the VLDB Endowment},
    pages = {622-633},
    title = {Scalable k-means++},
    year = {2012}
 }
 

Version:

$Revision: $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
CentroidSketch(weka.core.Instances initialSketch, weka.core.NormalizableDistance distanceFunction, int size, int seed) Constructor.

Method Summary

Modifier and Type	Method and Description
void	addReservoirToCurrentSketch() Add the reservoir to the current sketch.
void	aggregateReservoir(WeightedReservoirSample toAggregate) Aggregate the supplied reservoir into our reservoir.
double	distanceToSketch(weka.core.Instance toProcess) Computes the distance between the supplied instance and the current sketch.
weka.core.Instances	getCurrentSketch() Get the current sketch as a set of instances
weka.core.NormalizableDistance	getDistanceFunction() Get the distance function being used
WeightedReservoirSample	getReservoirSample() Get the reservoir sample
weka.core.TechnicalInformation	getTechnicalInformation()
java.lang.String	globalInfo() Overview information for this class
void	process(weka.core.Instance toProcess, boolean updateDistanceFunction) Processes an instance - basically updates the reservoir
void	resetReservoir() Clear the reservoir
void	setDistanceFunction(weka.core.NormalizableDistance distFunc) Set the distance function to use

Methods inherited from class java.lang.Object

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

Constructor Detail

CentroidSketch

public CentroidSketch(weka.core.Instances initialSketch,
                      weka.core.NormalizableDistance distanceFunction,
                      int size,
                      int seed)

Constructor.

Parameters:

initialSketch - the initial starting point (typically one randomly chosen instance for the k-means|| algorithm)

distanceFunction - the distance function to use

size - the size of the reservoir (i.e. how many points to consider adding to the sketch at each iteration)

seed - the seed for random number generation

Method Detail

globalInfo

public java.lang.String globalInfo()

Overview information for this class

Returns:

overview help information

process

public void process(weka.core.Instance toProcess,
                    boolean updateDistanceFunction)

Processes an instance - basically updates the reservoir

Parameters:

toProcess - the instance to process

updateDistanceFunction - true if we should update the distance function with this instance (i.e. update the ranges for numeric attributes)

distanceToSketch

public double distanceToSketch(weka.core.Instance toProcess)

Computes the distance between the supplied instance and the current sketch. This is the distance to the closest point in the sketch.

Parameters:

toProcess - the instance to process

Returns:

the distance to the current sketch

getDistanceFunction

public weka.core.NormalizableDistance getDistanceFunction()

Get the distance function being used

Returns:

the distance function

setDistanceFunction

public void setDistanceFunction(weka.core.NormalizableDistance distFunc)

Set the distance function to use

Parameters:

distFunc - the distance function to use

getReservoirSample

public WeightedReservoirSample getReservoirSample()

Get the reservoir sample

Returns:

the reservoir sample

getCurrentSketch

public weka.core.Instances getCurrentSketch()

Get the current sketch as a set of instances

Returns:

the current sketch

aggregateReservoir

public void aggregateReservoir(WeightedReservoirSample toAggregate)
                        throws java.lang.Exception

Aggregate the supplied reservoir into our reservoir. Does not increase the size of the sample.

Parameters:

toAggregate - the reservoir sample to aggregate

Throws:

java.lang.Exception - if the structure of the instances in the sample to aggregate does not match the structure of our sketch

resetReservoir

public void resetReservoir()

Clear the reservoir

addReservoirToCurrentSketch

public void addReservoirToCurrentSketch()
                                 throws java.lang.Exception

Add the reservoir to the current sketch. Also resets the reservoir.

Throws:

java.lang.Exception

getTechnicalInformation

public weka.core.TechnicalInformation getTechnicalInformation()

Specified by:

getTechnicalInformation in interface weka.core.TechnicalInformationHandler