org.deeplearning4j.nn.graph.vertex

Interface GraphVertex

All Superinterfaces:

Serializable, Trainable

All Known Implementing Classes:

BaseGraphVertex, BaseWrapperVertex, DuplicateToTimeSeriesVertex, ElementWiseVertex, FrozenVertex, InputVertex, L2NormalizeVertex, L2Vertex, LastTimeStepVertex, LayerVertex, MergeVertex, PoolHelperVertex, PreprocessorVertex, ReshapeVertex, ReverseTimeSeriesVertex, SameDiffGraphVertex, ScaleVertex, ShiftVertex, StackVertex, SubsetVertex, UnstackVertex

public interface GraphVertex
extends Trainable, Serializable

A GraphVertex is a vertex in the computation graph. It may contain Layer, or define some arbitrary forward/backward pass behaviour based on the inputs.
The purposes of GraphVertex instances are as follows: 1. To track the (local) network connection structure: i.e., a GraphVertex knows about the vertices on the input and output sides 2. To store intermediate results (activations and epsilons) 3. To allow forward pass and backward pass to be conducted, once the intermediate results are set

Author:

Alex Black

Method Summary

Modifier and Type	Method and Description
boolean	canDoBackward() Whether the GraphVertex can do backward pass.
boolean	canDoForward() Whether the GraphVertex can do forward pass.
void	clear() Clear the internal state (if any) of the GraphVertex.
void	clearVertex() This method clears inpjut for this vertex
Pair<Gradient,INDArray[]>	doBackward(boolean tbptt, LayerWorkspaceMgr workspaceMgr) Do backward pass
INDArray	doForward(boolean training, LayerWorkspaceMgr workspaceMgr) Do forward pass using the stored inputs
Pair<INDArray,MaskState>	feedForwardMaskArrays(INDArray[] maskArrays, MaskState currentMaskState, int minibatchSize)
INDArray	getEpsilon() Get the epsilon/error (i.e., dL/dOutput) array previously set for this GraphVertex
INDArray[]	getInputs() Get the array of inputs previously set for this GraphVertex
VertexIndices[]	getInputVertices() A representation of the vertices that are inputs to this vertex (inputs duing forward pass) Specifically, if inputVertices[X].getVertexIndex() = Y, and inputVertices[X].getVertexEdgeNumber() = Z then the Zth output connection (see getNumOutputConnections() of vertex Y is the Xth input to this vertex
Layer	getLayer() Get the Layer (if any).
int	getNumInputArrays() Get the number of input arrays.
int	getNumOutputConnections() Get the number of outgoing connections from this GraphVertex.
VertexIndices[]	getOutputVertices() A representation of the vertices that this vertex is connected to (outputs duing forward pass) Specifically, if outputVertices[X].getVertexIndex() = Y, and outputVertices[X].getVertexEdgeNumber() = Z then the Xth output of this vertex is connected to the Zth input of vertex Y
int	getVertexIndex() Get the index of the GraphVertex
String	getVertexName() Get the name/label of the GraphVertex
boolean	hasLayer() Whether the GraphVertex contains a Layer object or not
boolean	isInputVertex() Whether the GraphVertex is an input vertex
boolean	isOutputVertex() Whether the GraphVertex is an output vertex
Map<String,INDArray>	paramTable(boolean backpropOnly) Get the parameter table for the vertex
void	setBackpropGradientsViewArray(INDArray backpropGradientsViewArray) See Model.setBackpropGradientsViewArray(INDArray)
void	setEpsilon(INDArray epsilon) Set the errors (epsilon - aka dL/dActivation) for this GraphVertex
void	setInput(int inputNumber, INDArray input, LayerWorkspaceMgr workspaceMgr) Set the input activations.
void	setInputs(INDArray... inputs) Set all inputs for this GraphVertex
void	setInputVertices(VertexIndices[] inputVertices) Sets the input vertices.
void	setLayerAsFrozen() Only applies to layer vertices.
void	setOutputVertex(boolean outputVertex) Set the GraphVertex to be an output vertex
void	setOutputVertices(VertexIndices[] outputVertices) set the output vertices.

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

getConfig, getGradientsViewArray, numParams, params, updaterDivideByMinibatch

Method Detail

getVertexName

String getVertexName()

Get the name/label of the GraphVertex

getVertexIndex

int getVertexIndex()

Get the index of the GraphVertex

getNumInputArrays

int getNumInputArrays()

Get the number of input arrays. For example, a Layer may have only one input array, but in general a GraphVertex may have an arbtrary (>=1) number of input arrays (for example, from multiple other layers)

getNumOutputConnections

int getNumOutputConnections()

Get the number of outgoing connections from this GraphVertex. A GraphVertex may only have a single output (for example, the activations out of a layer), but this output may be used as the input to an arbitrary number of other GraphVertex instances. This method returns the number of GraphVertex instances the output of this GraphVertex is input for.

getInputVertices

VertexIndices[] getInputVertices()

A representation of the vertices that are inputs to this vertex (inputs duing forward pass)
Specifically, if inputVertices[X].getVertexIndex() = Y, and inputVertices[X].getVertexEdgeNumber() = Z then the Zth output connection (see getNumOutputConnections() of vertex Y is the Xth input to this vertex

setInputVertices

void setInputVertices(VertexIndices[] inputVertices)

Sets the input vertices.

See Also:

getInputVertices()

getOutputVertices

VertexIndices[] getOutputVertices()

A representation of the vertices that this vertex is connected to (outputs duing forward pass) Specifically, if outputVertices[X].getVertexIndex() = Y, and outputVertices[X].getVertexEdgeNumber() = Z then the Xth output of this vertex is connected to the Zth input of vertex Y

setOutputVertices

void setOutputVertices(VertexIndices[] outputVertices)

set the output vertices.

See Also:

getOutputVertices()

hasLayer

boolean hasLayer()

Whether the GraphVertex contains a Layer object or not

isInputVertex

boolean isInputVertex()

Whether the GraphVertex is an input vertex

isOutputVertex

boolean isOutputVertex()

Whether the GraphVertex is an output vertex

setOutputVertex

void setOutputVertex(boolean outputVertex)

Set the GraphVertex to be an output vertex

getLayer

Layer getLayer()

Get the Layer (if any). Returns null if hasLayer() == false

setInput

void setInput(int inputNumber,
              INDArray input,
              LayerWorkspaceMgr workspaceMgr)

Set the input activations.

Parameters:

inputNumber - Must be in range 0 to getNumInputArrays()-1

input - The input array

workspaceMgr -

setEpsilon

void setEpsilon(INDArray epsilon)

Set the errors (epsilon - aka dL/dActivation) for this GraphVertex

clear

void clear()

Clear the internal state (if any) of the GraphVertex. For example, any stored inputs/errors

canDoForward

boolean canDoForward()

Whether the GraphVertex can do forward pass. Typically, this is just whether all inputs are set.

canDoBackward

boolean canDoBackward()

Whether the GraphVertex can do backward pass. Typically, this is just whether all errors/epsilons are set

doForward

INDArray doForward(boolean training,
                   LayerWorkspaceMgr workspaceMgr)

Do forward pass using the stored inputs

Parameters:

training - if true: forward pass at training time. If false: forward pass at test time

Returns:

The output (for example, activations) of the GraphVertex

doBackward

Pair<Gradient,INDArray[]> doBackward(boolean tbptt,
                                     LayerWorkspaceMgr workspaceMgr)

Do backward pass

Parameters:

tbptt - If true: do backprop using truncated BPTT

Returns:

The gradients (may be null), and the errors/epsilons for all inputs to this GraphVertex

getInputs

INDArray[] getInputs()

Get the array of inputs previously set for this GraphVertex

getEpsilon

INDArray getEpsilon()

Get the epsilon/error (i.e., dL/dOutput) array previously set for this GraphVertex

setInputs

void setInputs(INDArray... inputs)

Set all inputs for this GraphVertex

See Also:

setInput(int, INDArray, LayerWorkspaceMgr)

setBackpropGradientsViewArray

void setBackpropGradientsViewArray(INDArray backpropGradientsViewArray)

See Model.setBackpropGradientsViewArray(INDArray)

Parameters:

backpropGradientsViewArray -

feedForwardMaskArrays

Pair<INDArray,MaskState> feedForwardMaskArrays(INDArray[] maskArrays,
                                               MaskState currentMaskState,
                                               int minibatchSize)

setLayerAsFrozen

void setLayerAsFrozen()

Only applies to layer vertices. Will throw exceptions on others. If applied to a layer vertex it will treat the parameters of the layer within it as constant. Activations through these will be calculated as they would as test time regardless of training mode

clearVertex

void clearVertex()

This method clears inpjut for this vertex

paramTable

Map<String,INDArray> paramTable(boolean backpropOnly)

Get the parameter table for the vertex

Specified by:

paramTable in interface Trainable

Parameters:

backpropOnly - If true: exclude unsupervised training parameters

Returns:

Parameter table