org.deeplearning4j.models.embeddings.learning.impl.elements

Class GloVe<T extends SequenceElement>

java.lang.Object

org.deeplearning4j.models.embeddings.learning.impl.elements.GloVe<T>

All Implemented Interfaces:

ElementsLearningAlgorithm<T>

public class GloVe<T extends SequenceElement>
extends Object
implements ElementsLearningAlgorithm<T>

GloVe LearningAlgorithm implementation for SequenceVectors

Author:

[email protected]

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	GloVe.Builder<T extends SequenceElement>

Field Summary

Fields

Modifier and Type	Field and Description
protected double	alpha
protected int	batchSize
protected double	learningRate
protected int	maxmemory

Constructor Summary

Constructors

Constructor and Description
GloVe()

Method Summary

Modifier and Type	Method and Description
void	configure(VocabCache<T> vocabCache, WeightLookupTable<T> lookupTable, VectorsConfiguration configuration)
void	finish()
String	getCodeName()
boolean	isEarlyTerminationHit() Since GloVe is learning representations using elements CoOccurences, all training is done in GloVe class internally, so only first thread will execute learning process, and the rest of parent threads will just exit learning process
double	learnSequence(Sequence<T> sequence, AtomicLong nextRandom, double learningRate) Learns sequence using GloVe algorithm
void	pretrain(SequenceIterator<T> iterator) pretrain is used to build CoOccurrence matrix for GloVe algorithm

Methods inherited from class java.lang.Object

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

Field Detail

alpha

protected double alpha

learningRate

protected double learningRate

maxmemory

protected int maxmemory

batchSize

protected int batchSize

Constructor Detail

GloVe

public GloVe()

Method Detail

getCodeName

public String getCodeName()

Specified by:

getCodeName in interface ElementsLearningAlgorithm<T extends SequenceElement>

finish

public void finish()

Specified by:

finish in interface ElementsLearningAlgorithm<T extends SequenceElement>

configure

public void configure(@NonNull
                      VocabCache<T> vocabCache,
                      @NonNull
                      WeightLookupTable<T> lookupTable,
                      @NonNull
                      VectorsConfiguration configuration)

Specified by:

configure in interface ElementsLearningAlgorithm<T extends SequenceElement>

pretrain

public void pretrain(@NonNull
                     SequenceIterator<T> iterator)

pretrain is used to build CoOccurrence matrix for GloVe algorithm

Specified by:

pretrain in interface ElementsLearningAlgorithm<T extends SequenceElement>

Parameters:

iterator -

learnSequence

public double learnSequence(@NonNull
                            Sequence<T> sequence,
                            @NonNull
                            AtomicLong nextRandom,
                            double learningRate)

Learns sequence using GloVe algorithm

Specified by:

learnSequence in interface ElementsLearningAlgorithm<T extends SequenceElement>

Parameters:

sequence -

nextRandom -

learningRate -

Returns:

average score for this sequence

isEarlyTerminationHit

public boolean isEarlyTerminationHit()

Since GloVe is learning representations using elements CoOccurences, all training is done in GloVe class internally, so only first thread will execute learning process, and the rest of parent threads will just exit learning process

Specified by:

isEarlyTerminationHit in interface ElementsLearningAlgorithm<T extends SequenceElement>

Returns:

True, if training should stop, False otherwise.