oracle.spatial.hadoop.vector.cluster.kmeans

Interface CriterionFunction

All Known Implementing Classes:

EuclideanDistanceCriterionFunction, SquaredErrorCriterionFunction

public interface CriterionFunction

Represents a criterion function used to determine if a cluster algorithm has finished during the cluster center update phase at reduce time. The life cycle of a an instace of this type is as follows: setup() //once per instance beforeClusterCenterUpdated() //once per cluster addClusterMember() //one or more times per cluster clusterCenterUpdated() //once per cluster getCriterionFunctionValue() // one or more times per cluster once clusterCenterUpdated() has been called isClusterConverging() // one or more times per cluster once clusterCenterUpdated() has been called

Author:

hsaucedo

Method Summary

Modifier and Type	Method and Description
void	addClusterMember(RecordInfo member, double[] memberPoint) Adds the data and point of an individual cluster member.
void	beforeClusterCenterUpdated(int clusterId, double[] clusterCenter, double critFunValue, long memberCount) Sets the current cluster and its state before the cluster point is updated.
void	clusterCenterUpdated(double[] clusterCenter) Sets the updated cluster center point
double	getCriterionFunctionValue() Gets the new criterion function value calculated using the updated cluster center point
boolean	isClusterConverging() Tell whether the criterion function is converging for the current cluster
void	setup(Configuration conf) Perform initialization tasks.

Method Detail

addClusterMember

void addClusterMember(RecordInfo member,
                      double[] memberPoint)

Adds the data and point of an individual cluster member. The implementer may need to copy the parameters values as the references could be reused.

Parameters:

member - a cluster member as a RecordInfo instance

memberPoint - a cluster member point

beforeClusterCenterUpdated

void beforeClusterCenterUpdated(int clusterId,
                                double[] clusterCenter,
                                double critFunValue,
                                long memberCount)

Sets the current cluster and its state before the cluster point is updated.

Parameters:

clusterId - the cluster id

clusterCenter - the cluster center point before updated

critFunValue - the last criterion function value

memberCount - the number of cluster members

clusterCenterUpdated

void clusterCenterUpdated(double[] clusterCenter)

Sets the updated cluster center point

Parameters:

clusterCenter - the cluster center point ordinates

getCriterionFunctionValue

double getCriterionFunctionValue()

Gets the new criterion function value calculated using the updated cluster center point

Returns:

the newest criterion function value

isClusterConverging

boolean isClusterConverging()

Tell whether the criterion function is converging for the current cluster

Returns:

true if the function is converging for the current cluster

setup

void setup(Configuration conf)

Perform initialization tasks. Called once per instance

Parameters:

conf - the job configuration