oracle.spatial.hadoop.vector.cluster.kmeans

Class KMeansSampler<K,V>

java.lang.Object

oracle.spatial.hadoop.vector.cluster.kmeans.KMeansSampler<K,V>

Type Parameters:

K - The type of the keys

V - The type of the vaues

public class KMeansSampler<K,V>
extends java.lang.Object

Samples a dataset to find the initial k cluster centers

Author:

hsaucedo

Constructor Summary

Constructors

Constructor and Description
KMeansSampler(java.lang.Class<? extends <any>> inputFormatClass, java.lang.Class<? extends RecordInfoProvider<K,V>> riProviderClass, int dimensions) Creates a new instance

Method Summary

Modifier and Type	Method and Description
double[]	sampleClusterCenters(int k, int maxSampleRecords, int maxSamplePartitions, JobConf conf) Samples the input data and return the k initial cluster center points
double[]	sampleClusterCenters(int k, JobConf conf) Samples the input data and return the k initial cluster center points
void	writeClusterCenters(Path outFile, int k, int maxSampleRecords, int maxSamplePartitions, JobConf conf) Samples the input data to find the initial cluster centers and write them to the given file

Methods inherited from class java.lang.Object

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

Constructor Detail

KMeansSampler

public KMeansSampler(java.lang.Class<? extends <any>> inputFormatClass,
                     java.lang.Class<? extends RecordInfoProvider<K,V>> riProviderClass,
                     int dimensions)

Creates a new instance

Parameters:

inputFormatClass - InputFormat used to read the data

riProviderClass - RecordInfoProvider implementation used to extract spatial information from the records

dimensions -

Method Detail

sampleClusterCenters

public double[] sampleClusterCenters(int k,
                                     int maxSampleRecords,
                                     int maxSamplePartitions,
                                     JobConf conf)
                              throws java.lang.Exception

Samples the input data and return the k initial cluster center points

Parameters:

k - the number of initial clusters

maxSampleRecords - the maximum number of records that should be sampled

maxSamplePartitions - the maximum number of partitions to be sampled

conf - the job configuration

Returns:

the initial cluster center point ordinates

Throws:

java.lang.Exception

sampleClusterCenters

public double[] sampleClusterCenters(int k,
                                     JobConf conf)
                              throws java.lang.Exception

Samples the input data and return the k initial cluster center points

Parameters:

k - the number of initial clusters

conf - the job configuration

Returns:

the initial cluster center point ordinates

Throws:

java.lang.Exception

writeClusterCenters

public void writeClusterCenters(Path outFile,
                                int k,
                                int maxSampleRecords,
                                int maxSamplePartitions,
                                JobConf conf)
                         throws java.lang.Exception

Samples the input data to find the initial cluster centers and write them to the given file

Parameters:

outFile - the path where the initial clusters will be written to

k - the number of initial clusters

maxSampleRecords - the maximum number of records that should be sampled

maxSamplePartitions - the maximum number of partitions to be sampled

conf - the job configuration

Throws:

java.lang.Exception