org.deeplearning4j.earlystopping.scorecalc

Class VAEReconErrorScoreCalculator

java.lang.Object

org.deeplearning4j.earlystopping.scorecalc.base.BaseScoreCalculator<Model>

org.deeplearning4j.earlystopping.scorecalc.VAEReconErrorScoreCalculator

All Implemented Interfaces:

Serializable, ScoreCalculator<Model>

public class VAEReconErrorScoreCalculator
extends BaseScoreCalculator<Model>

Score function for variational autoencoder reconstruction error for a MultiLayerNetwork or ComputationGraph.
VariationalAutoencoder layer must be first layer in the network

See Also:

for reconstruction probability, Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected RegressionEvaluation	evaluation
protected RegressionEvaluation.Metric	metric

Fields inherited from class org.deeplearning4j.earlystopping.scorecalc.base.BaseScoreCalculator

exampleCount, iterator, mdsIterator, minibatchCount, scoreSum

Constructor Summary

Constructors

Constructor and Description
VAEReconErrorScoreCalculator(RegressionEvaluation.Metric metric, DataSetIterator iterator) Constructor for reconstruction *ERROR*

Method Summary

Modifier and Type	Method and Description
protected double	finalScore(double scoreSum, int minibatchCount, int exampleCount)
boolean	minimizeScore()
protected INDArray[]	output(Model network, INDArray[] input, INDArray[] fMask, INDArray[] lMask)
protected INDArray	output(Model net, INDArray input, INDArray fMask, INDArray lMask)
protected void	reset()
protected double	scoreMinibatch(Model network, INDArray[] features, INDArray[] labels, INDArray[] fMask, INDArray[] lMask, INDArray[] output)
protected double	scoreMinibatch(Model network, INDArray features, INDArray labels, INDArray fMask, INDArray lMask, INDArray output)

Methods inherited from class org.deeplearning4j.earlystopping.scorecalc.base.BaseScoreCalculator

arr, calculateScore, get0

Methods inherited from class java.lang.Object

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

Field Detail

metric

protected final RegressionEvaluation.Metric metric

evaluation

protected RegressionEvaluation evaluation

Constructor Detail

VAEReconErrorScoreCalculator

public VAEReconErrorScoreCalculator(RegressionEvaluation.Metric metric,
                                    DataSetIterator iterator)

Constructor for reconstruction *ERROR*

Parameters:

metric -

iterator -

Method Detail

reset

protected void reset()

Specified by:

reset in class BaseScoreCalculator<Model>

output

protected INDArray output(Model net,
                          INDArray input,
                          INDArray fMask,
                          INDArray lMask)

Specified by:

output in class BaseScoreCalculator<Model>

output

protected INDArray[] output(Model network,
                            INDArray[] input,
                            INDArray[] fMask,
                            INDArray[] lMask)

Specified by:

output in class BaseScoreCalculator<Model>

scoreMinibatch

protected double scoreMinibatch(Model network,
                                INDArray features,
                                INDArray labels,
                                INDArray fMask,
                                INDArray lMask,
                                INDArray output)

Overrides:

scoreMinibatch in class BaseScoreCalculator<Model>

scoreMinibatch

protected double scoreMinibatch(Model network,
                                INDArray[] features,
                                INDArray[] labels,
                                INDArray[] fMask,
                                INDArray[] lMask,
                                INDArray[] output)

Specified by:

scoreMinibatch in class BaseScoreCalculator<Model>

finalScore

protected double finalScore(double scoreSum,
                            int minibatchCount,
                            int exampleCount)

Specified by:

finalScore in class BaseScoreCalculator<Model>

minimizeScore

public boolean minimizeScore()

Returns:

If true: the score should be minimized. If false: the score should be maximized.