org.deeplearning4j.models.sequencevectors

Class SequenceVectors<T extends SequenceElement>

java.lang.Object

org.deeplearning4j.models.embeddings.wordvectors.WordVectorsImpl<T>

org.deeplearning4j.models.sequencevectors.SequenceVectors<T>

All Implemented Interfaces:

Serializable, WordVectors

Direct Known Subclasses:

Glove, Node2Vec, Word2Vec

public class SequenceVectors<T extends SequenceElement>
extends WordVectorsImpl<T>
implements WordVectors

SequenceVectors implements abstract features extraction for Sequences and SequenceElements, using SkipGram, CBOW or DBOW (for Sequence features extraction).

Author:

[email protected]

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
protected class	SequenceVectors.AsyncSequencer This class is used to fetch data from iterator in background thread, and convert it to List It becomes very usefull if text processing pipeline behind iterator is complex, and we're not loading data from simple text file with whitespaces as separator.
static class	SequenceVectors.Builder<T extends SequenceElement>

Field Summary

Fields

Modifier and Type	Field and Description
protected VectorsConfiguration	configuration
protected boolean	configured
protected ElementsLearningAlgorithm<T>	elementsLearningAlgorithm
protected boolean	enableScavenger
protected Set<VectorsListener<T>>	eventListeners
protected WordVectors	existingModel
protected SequenceIterator<T>	iterator
protected static org.slf4j.Logger	log
protected com.google.common.util.concurrent.AtomicDouble	scoreElements
protected com.google.common.util.concurrent.AtomicDouble	scoreSequences
protected SequenceLearningAlgorithm<T>	sequenceLearningAlgorithm
protected T	unknownElement
protected int	vocabLimit

Fields inherited from class org.deeplearning4j.models.embeddings.wordvectors.WordVectorsImpl

batchSize, DEFAULT_UNK, layerSize, learningRate, learningRateDecayWords, lookupTable, minLearningRate, minWordFrequency, modelUtils, negative, numEpochs, numIterations, resetModel, sampling, seed, stopWords, trainElementsVectors, trainSequenceVectors, useAdeGrad, useUnknown, variableWindows, vocab, window, workers

Constructor Summary

Constructors

Constructor and Description
SequenceVectors()

Method Summary

Modifier and Type	Method and Description
void	buildVocab() Builds vocabulary from provided SequenceIterator instance
void	fit() Starts training over
double	getElementsScore()
double	getSequencesScore()
String	getUNK()
org.nd4j.linalg.api.ndarray.INDArray	getWordVectorMatrix(String word) Get the word vector for a given matrix
protected void	initLearners()
void	setUNK(String UNK)
protected void	trainSequence(Sequence<T> sequence, AtomicLong nextRandom, double alpha)

Methods inherited from class org.deeplearning4j.models.embeddings.wordvectors.WordVectorsImpl

accuracy, getLayerSize, getWordVector, getWordVectorMatrixNormalized, getWordVectors, getWordVectorsMean, hasWord, indexOf, lookupTable, setLookupTable, setModelUtils, setVocab, similarity, similarWordsInVocabTo, update, update, vocab, wordsNearest, wordsNearest, wordsNearest, wordsNearestSum, wordsNearestSum, wordsNearestSum

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.deeplearning4j.models.embeddings.wordvectors.WordVectors

accuracy, getWordVector, getWordVectorMatrixNormalized, getWordVectors, getWordVectorsMean, hasWord, indexOf, lookupTable, setModelUtils, similarity, similarWordsInVocabTo, vocab, wordsNearest, wordsNearest, wordsNearest, wordsNearestSum, wordsNearestSum, wordsNearestSum

Field Detail

iterator

protected transient SequenceIterator<T extends SequenceElement> iterator

elementsLearningAlgorithm

protected transient ElementsLearningAlgorithm<T extends SequenceElement> elementsLearningAlgorithm

sequenceLearningAlgorithm

protected transient SequenceLearningAlgorithm<T extends SequenceElement> sequenceLearningAlgorithm

configuration

protected VectorsConfiguration configuration

log

protected static final org.slf4j.Logger log

existingModel

protected transient WordVectors existingModel

unknownElement

protected transient T extends SequenceElement unknownElement

scoreElements

protected transient com.google.common.util.concurrent.AtomicDouble scoreElements

scoreSequences

protected transient com.google.common.util.concurrent.AtomicDouble scoreSequences

configured

protected transient boolean configured

enableScavenger

protected boolean enableScavenger

vocabLimit

protected int vocabLimit

eventListeners

protected transient Set<VectorsListener<T extends SequenceElement>> eventListeners

Constructor Detail

SequenceVectors

public SequenceVectors()

Method Detail

getUNK

public String getUNK()

Specified by:

getUNK in interface WordVectors

setUNK

public void setUNK(String UNK)

Specified by:

setUNK in interface WordVectors

getElementsScore

public double getElementsScore()

getSequencesScore

public double getSequencesScore()

getWordVectorMatrix

public org.nd4j.linalg.api.ndarray.INDArray getWordVectorMatrix(String word)

Description copied from interface: WordVectors

Get the word vector for a given matrix

Specified by:

getWordVectorMatrix in interface WordVectors

Overrides:

getWordVectorMatrix in class WordVectorsImpl<T extends SequenceElement>

Parameters:

word - the word to get the matrix for

Returns:

the ndarray for this word

buildVocab

public void buildVocab()

Builds vocabulary from provided SequenceIterator instance

initLearners

protected void initLearners()

fit

public void fit()

Starts training over

trainSequence

protected void trainSequence(@NonNull
                             Sequence<T> sequence,
                             AtomicLong nextRandom,
                             double alpha)