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

Class GloVe.Builder<T extends SequenceElement>

java.lang.Object

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

Enclosing class:

GloVe<T extends SequenceElement>

public static class GloVe.Builder<T extends SequenceElement>
extends Object

Field Summary

Fields

Modifier and Type	Field and Description
protected double	alpha
protected int	batchSize
protected double	learningRate
protected int	maxmemory
protected boolean	shuffle
protected boolean	symmetric
protected double	xMax

Constructor Summary

Constructors

Constructor and Description
Builder()

Method Summary

Modifier and Type	Method and Description
GloVe.Builder<T>	alpha(double alpha) Parameter in exponent of weighting function; default 0.75
GloVe.Builder<T>	batchSize(int batchSize) This parameter specifies batch size for each thread.
GloVe<T>	build()
GloVe.Builder<T>	learningRate(double eta) Initial learning rate; default 0.05
GloVe.Builder<T>	maxMemory(int gbytes) This method allows you to specify maximum memory available for CoOccurrence map builder.
GloVe.Builder<T>	shuffle(boolean reallyShuffle) Parameter specifying, if cooccurrences list should be shuffled between training epochs
GloVe.Builder<T>	symmetric(boolean reallySymmetric) Parameters specifying, if cooccurrences list should be build into both directions from any current word.
GloVe.Builder<T>	xMax(double xMax) Parameter specifying cutoff in weighting function; default 100.0

Methods inherited from class java.lang.Object

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

Field Detail

xMax

protected double xMax

alpha

protected double alpha

learningRate

protected double learningRate

shuffle

protected boolean shuffle

symmetric

protected boolean symmetric

maxmemory

protected int maxmemory

batchSize

protected int batchSize

Constructor Detail

Builder

public Builder()

Method Detail

batchSize

public GloVe.Builder<T> batchSize(int batchSize)

This parameter specifies batch size for each thread. Also, if shuffle == TRUE, this batch will be shuffled before processing. Default value: 1000;

Parameters:

batchSize -

Returns:

learningRate

public GloVe.Builder<T> learningRate(double eta)

Initial learning rate; default 0.05

Parameters:

eta -

Returns:

alpha

public GloVe.Builder<T> alpha(double alpha)

Parameter in exponent of weighting function; default 0.75

Parameters:

alpha -

Returns:

maxMemory

public GloVe.Builder<T> maxMemory(int gbytes)

This method allows you to specify maximum memory available for CoOccurrence map builder. Please note: this option can be considered a debugging method. In most cases setting proper -Xmx argument set to JVM is enough to limit this algorithm. Please note: this option won't override -Xmx JVM value.

Parameters:

gbytes - memory limit, in gigabytes

Returns:

xMax

public GloVe.Builder<T> xMax(double xMax)

Parameter specifying cutoff in weighting function; default 100.0

Parameters:

xMax -

Returns:

shuffle

public GloVe.Builder<T> shuffle(boolean reallyShuffle)

Parameter specifying, if cooccurrences list should be shuffled between training epochs

Parameters:

reallyShuffle -

Returns:

symmetric

public GloVe.Builder<T> symmetric(boolean reallySymmetric)

Parameters specifying, if cooccurrences list should be build into both directions from any current word.

Parameters:

reallySymmetric -

Returns:

build

public GloVe<T> build()