org.deeplearning4j.nn.conf

Class ComputationGraphConfiguration

java.lang.Object

org.deeplearning4j.nn.conf.ComputationGraphConfiguration

All Implemented Interfaces:

Serializable, Cloneable

public class ComputationGraphConfiguration
extends Object
implements Serializable, Cloneable

ComputationGraphConfiguration is a configuration object for neural networks with arbitrary connection structure. It is analogous to MultiLayerConfiguration, but allows considerably greater flexibility for the network architecture.
Specifically, the network architecture is a directed acyclic graph, where each vertex in the graph is a GraphVertex, which may for example be a layer or a vertex/object that defines arbitrary forward and backward pass functionality.
Note that the ComputationGraph may have an arbitrary number of inputs (multiple independent inputs, possibly of different types), and an arbitrary number of outputs (for example, multiple OutputLayer instances. Typical usage:
ComputationGraphConfiguration conf = new NeuralNetConfiguration.Builder()....graphBuilder()...build();

Author:

Alex Black

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	ComputationGraphConfiguration.GraphBuilder

Field Summary

Fields

Modifier and Type	Field and Description
protected boolean	backprop
protected BackpropType	backpropType
protected NeuralNetConfiguration	defaultConfiguration
protected List<String>	networkInputs List of inputs to the network, by name
protected List<String>	networkOutputs List of network outputs, by name
protected boolean	pretrain
protected int	tbpttBackLength
protected int	tbpttFwdLength
protected Map<String,List<String>>	vertexInputs
protected Map<String,GraphVertex>	vertices

Constructor Summary

Constructors

Constructor and Description
ComputationGraphConfiguration()

Method Summary

Modifier and Type	Method and Description
void	addPreProcessors(InputType... inputTypes) Add preprocessors automatically, given the specified types of inputs for the network.
ComputationGraphConfiguration	clone()
static ComputationGraphConfiguration	fromJson(String json) Create a computation graph configuration from json
static ComputationGraphConfiguration	fromYaml(String json) Create a neural net configuration from json
String	toJson()
String	toString()
String	toYaml()
void	validate() Check the configuration, make sure it is valid

Methods inherited from class java.lang.Object

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

Field Detail

vertices

protected Map<String,GraphVertex> vertices

vertexInputs

protected Map<String,List<String>> vertexInputs

networkInputs

protected List<String> networkInputs

List of inputs to the network, by name

networkOutputs

protected List<String> networkOutputs

List of network outputs, by name

pretrain

protected boolean pretrain

backprop

protected boolean backprop

backpropType

protected BackpropType backpropType

tbpttFwdLength

protected int tbpttFwdLength

tbpttBackLength

protected int tbpttBackLength

defaultConfiguration

protected NeuralNetConfiguration defaultConfiguration

Constructor Detail

ComputationGraphConfiguration

public ComputationGraphConfiguration()

Method Detail

toYaml

public String toYaml()

Returns:

JSON representation of configuration

fromYaml

public static ComputationGraphConfiguration fromYaml(String json)

Create a neural net configuration from json

Parameters:

json - the neural net configuration from json

Returns:

ComputationGraphConfiguration

toJson

public String toJson()

Returns:

JSON representation of computation graph configuration

fromJson

public static ComputationGraphConfiguration fromJson(String json)

Create a computation graph configuration from json

Parameters:

json - the neural net configuration from json

Returns:

ComputationGraphConfiguration

toString

public String toString()

Overrides:

toString in class Object

clone

public ComputationGraphConfiguration clone()

Overrides:

clone in class Object

validate

public void validate()

Check the configuration, make sure it is valid

Throws:

IllegalStateException - if configuration is not valid

addPreProcessors

public void addPreProcessors(InputType... inputTypes)

Add preprocessors automatically, given the specified types of inputs for the network. Inputs are specified using the InputType class, in the same order in which the inputs were defined in the original configuration.
For example, in a network with two inputs: a convolutional input (28x28x1 images) and feed forward inputs, use .addPreProcessors(InputType.convolutional(1,28,28),InputType.feedForward()).
For the CNN->Dense and CNN->RNN transitions, the nIns on the Dense/RNN layers will also be added automatically. NOTE: This method will be called automatically when using the ComputationGraphConfiguration.GraphBuilder.setInputTypes(InputType...) functionality. See that method for details.