net.sourceforge.cilib.clustering.kmeans
Interface CentroidsInitialisationStrategy

All Superinterfaces:

Cloneable, Serializable

All Known Implementing Classes:

DataSetBasedCentroidsInitialisationStrategy, RandomCentroidsInitialisationStrategy

public interface CentroidsInitialisationStrategy
extends Serializable, Cloneable

This strategy allows for different ways of initializing the centroids of a clustering. The two main approaches are either:

• Randomly (RandomCentroidsInitialisationStrategy); or
• Based on random patterns chosen from the dataset (DataSetBasedCentroidsInitialisationStrategy)

Author:

Theuns Cloete

Method Summary

Vector

initialise(ClusteringProblem problem, ClusterableDataSet dataset) Initialize the centroid vectors for a clustering.

Methods inherited from interface net.sourceforge.cilib.util.Cloneable

getClone

Method Detail

initialise

Vector initialise(ClusteringProblem problem,
                  ClusterableDataSet dataset)

Initialize the centroid vectors for a clustering. Each centroid is appended to a Vector that represents all the centroids. This Vector is then returned. The problem and/or dataset that are currently being clustered can be used to get information about the clustering, such as the dimension of the search space and centroids.

Parameters:

problem - the ClusteringProblem currently being optimized

dataset - the ClusterableDataSet currently being clustered

Returns:

a Vector that represents all the centroids