edu.umass.cs.mallet.projects.seg_plus_coref.condclust.pipe.iterator

Class NodeClusterPairIterator

java.lang.Object

edu.umass.cs.mallet.base.pipe.iterator.AbstractPipeInputIterator

edu.umass.cs.mallet.projects.seg_plus_coref.condclust.pipe.iterator.NodeClusterPairIterator

All Implemented Interfaces:

PipeInputIterator, Iterator

public class NodeClusterPairIterator
extends AbstractPipeInputIterator

Generates instances of NodeClusterPairs by one of two ways: (1) randomly clustering the data following the Chinese-restaurant style generative process, or (2) given the true clustering, generate NodeClusterPairs by sampling from the true clusters. Assumes for now that all clusters are pure (i.e. always make correct clustering decisions).

Field Summary

Fields inherited from class edu.umass.cs.mallet.base.pipe.iterator.AbstractPipeInputIterator

parentInstance

Constructor Summary

Constructors

Constructor and Description
NodeClusterPairIterator(Collection clusters, Random r, double positiveInstanceRatio, boolean generateSampledInstances)
NodeClusterPairIterator(Collection clusters, Random r, double positiveInstanceRatio, boolean generateSampledInstances, int sampleSize) Randomly choose nodes and build clusters.
NodeClusterPairIterator(List data)

Method Summary

Modifier and Type	Method and Description
boolean	hasNext()
Object	next()
Instance	nextInstance()
void	remove()

Methods inherited from class edu.umass.cs.mallet.base.pipe.iterator.AbstractPipeInputIterator

setParentInstance

Methods inherited from class java.lang.Object

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

Methods inherited from interface java.util.Iterator

forEachRemaining

Constructor Detail

NodeClusterPairIterator

public NodeClusterPairIterator(Collection clusters,
                               Random r,
                               double positiveInstanceRatio,
                               boolean generateSampledInstances,
                               int sampleSize)

Randomly choose nodes and build clusters. Each time we must decide whether a node belongs in an existing cluster, we make a nodeClusterPair instance.

Parameters:

clusters - true clustering

r - for randomly selecting nodes to cluster

negativeInstanceFrequency - include negative instances with frequency 1/negativeInstanceFrequency

generateSampledInstances - sample positive instances from true clusters by sampling

sampleSize - number of training instances

NodeClusterPairIterator

public NodeClusterPairIterator(Collection clusters,
                               Random r,
                               double positiveInstanceRatio,
                               boolean generateSampledInstances)

NodeClusterPairIterator

public NodeClusterPairIterator(List data)

Method Detail

hasNext

public boolean hasNext()

Specified by:

hasNext in interface Iterator

Specified by:

hasNext in class AbstractPipeInputIterator

nextInstance

public Instance nextInstance()

Specified by:

nextInstance in interface PipeInputIterator

Specified by:

nextInstance in class AbstractPipeInputIterator

next

public Object next()

Specified by:

next in interface Iterator

Overrides:

next in class AbstractPipeInputIterator

remove

public void remove()

Specified by:

remove in interface Iterator

Overrides:

remove in class AbstractPipeInputIterator