org.deeplearning4j.nn.conf.layers.samediff

Class SameDiffOutputLayer

java.lang.Object

org.deeplearning4j.nn.conf.layers.Layer

org.deeplearning4j.nn.conf.layers.samediff.AbstractSameDiffLayer

org.deeplearning4j.nn.conf.layers.samediff.SameDiffOutputLayer

All Implemented Interfaces:

Serializable, Cloneable, TrainingConfig

public abstract class SameDiffOutputLayer
extends AbstractSameDiffLayer

See Also:

Serialized Form

Nested Class Summary

Nested classes/interfaces inherited from class org.deeplearning4j.nn.conf.layers.samediff.AbstractSameDiffLayer

AbstractSameDiffLayer.Builder<T extends AbstractSameDiffLayer.Builder<T>>

Field Summary

Fields inherited from class org.deeplearning4j.nn.conf.layers.samediff.AbstractSameDiffLayer

biasUpdater, gradientNormalization, gradientNormalizationThreshold, regularization, regularizationBias, updater

Fields inherited from class org.deeplearning4j.nn.conf.layers.Layer

constraints, iDropout, layerName

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	SameDiffOutputLayer()

Method Summary

Modifier and Type	Method and Description
abstract String	activationsVertexName() Output layers should terminate in a single scalar value (i.e., a score) - however, sometimes the output activations (such as softmax probabilities) need to be returned.
abstract SDVariable	defineLayer(SameDiff sameDiff, SDVariable layerInput, SDVariable labels, Map<String,SDVariable> paramTable) Define the output layer
Layer	instantiate(NeuralNetConfiguration conf, Collection<TrainingListener> trainingListeners, int layerIndex, INDArray layerParamsView, boolean initializeParams, DataType networkDataType)
boolean	labelsRequired() Whether labels are required for calculating the score.

Methods inherited from class org.deeplearning4j.nn.conf.layers.samediff.AbstractSameDiffLayer

applyGlobalConfig, applyGlobalConfigToLayer, defineParameters, getLayerParams, getMemoryReport, getPreProcessorForInputType, getRegularizationByParam, getUpdaterByParam, initializeParameters, initializer, initWeights, isPretrainParam, onesMaskForInput, paramReshapeOrder, setNIn

Methods inherited from class org.deeplearning4j.nn.conf.layers.Layer

clone, getOutputType, initializeConstraints, resetLayerDefaultConfig, setDataType

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.deeplearning4j.nn.api.TrainingConfig

getGradientNormalization, getGradientNormalizationThreshold, getLayerName

Constructor Detail

SameDiffOutputLayer

protected SameDiffOutputLayer()

Method Detail

defineLayer

public abstract SDVariable defineLayer(SameDiff sameDiff,
                                       SDVariable layerInput,
                                       SDVariable labels,
                                       Map<String,SDVariable> paramTable)

Define the output layer

Parameters:

sameDiff - SameDiff instance

layerInput - Input to the layer

labels - Labels variable (or null if labelsRequired() returns false

paramTable - Parameter table - keys as defined by AbstractSameDiffLayer.defineParameters(SDLayerParams)

Returns:

The final layer variable corresponding to the score/loss during forward pass. This must be a single scalar value.

activationsVertexName

public abstract String activationsVertexName()

Output layers should terminate in a single scalar value (i.e., a score) - however, sometimes the output activations (such as softmax probabilities) need to be returned. When this is the case, we need to know the name of the SDVariable that corresponds to these.
If the final network activations are just the input to the layer, simply return "input"

Returns:

The name of the activations to return when performing forward pass

labelsRequired

public boolean labelsRequired()

Whether labels are required for calculating the score. Defaults to true - however, if the score can be calculated without labels (for example, in some output layers used for unsupervised learning) this can be set to false.

Returns:

True if labels are required to calculate the score/output, false otherwise.

instantiate

public Layer instantiate(NeuralNetConfiguration conf,
                         Collection<TrainingListener> trainingListeners,
                         int layerIndex,
                         INDArray layerParamsView,
                         boolean initializeParams,
                         DataType networkDataType)

Specified by:

instantiate in class AbstractSameDiffLayer