moa.clusterers.outliers.AnyOut.util

Class EMProjectedClustering

java.lang.Object

moa.clusterers.outliers.AnyOut.util.EMProjectedClustering

public class EMProjectedClustering
extends Object

Implements clustering via Expectation Maximization but return a clear partitioning of the data, i.e. a point belongs to only one cluster (the most probable)

Author:

kranen

Constructor Summary

Constructors

Constructor and Description
EMProjectedClustering()

Method Summary

Methods

Modifier and Type	Method and Description
int[][]	getEMClusteringVariances(double[][] pointArray, int k) Performs an EM clustering on the provided data set !! Only the variances are calculated and used for point assignments ! !!! the number k' of returned clusters might be smaller than k !!!
int[][]	getEMClusteringVariancesBestChoice(double[][] pointArray, int k, int nrOfChoices)

Methods inherited from class java.lang.Object

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

Constructor Detail

EMProjectedClustering

public EMProjectedClustering()

Method Detail

getEMClusteringVariances

public int[][] getEMClusteringVariances(double[][] pointArray,
                               int k)

Performs an EM clustering on the provided data set !! Only the variances are calculated and used for point assignments ! !!! the number k' of returned clusters might be smaller than k !!!

Parameters:

pointArray - the data set as an array[n][d] of n points with d dimensions

k - the number of requested partitions (!might return less)

Returns:

a mapping int[n][k'] of the n given points to the k' resulting clusters

getEMClusteringVariancesBestChoice

public int[][] getEMClusteringVariancesBestChoice(double[][] pointArray,
                                         int k,
                                         int nrOfChoices)