org.encog.neural.networks.layers
Class ContextLayer

java.lang.Object
  org.encog.neural.networks.layers.BasicLayer
      org.encog.neural.networks.layers.ContextLayer

All Implemented Interfaces:

java.io.Serializable, java.lang.Comparable<Layer>, ContextClearable, Layer, EncogPersistedObject

public class ContextLayer
extends BasicLayer
implements ContextClearable

Implements a context layer. A context layer is used to implement a simple recurrent neural network, such as an Elman or Jordan neural network. The context layer has a short-term memory. The context layer accept input, and provide the same data as output on the next cycle. This continues, and the context layer's output "one step" out of sync with the input.

Author:

jheaton

See Also:

Serialized Form

Constructor Summary

ContextLayer()
Default constructor, mainly so the workbench can easily create a default layer.

ContextLayer(ActivationFunction thresholdFunction, boolean hasThreshold, int neuronCount)
Construct a context layer with the parameters specified.

ContextLayer(int neuronCount)
Construct a default context layer that has the TANH activation function and the specified number of neurons.

Method Summary

void	clearContext() Reset the context values back to zero.
Persistor	createPersistor() Create a persistor for this layer.
NeuralData	getContext()
void	process(NeuralData pattern) Called to process input from the previous layer.
NeuralData	recur() Called to get the output from this layer when called in a recurrent manor.

Methods inherited from class org.encog.neural.networks.layers.BasicLayer

addNext, addNext, addSynapse, compareTo, compute, getActivationFunction, getDescription, getID, getName, getNetwork, getNeuronCount, getNext, getNextLayers, getThreshold, getThreshold, getX, getY, hasThreshold, isConnectedTo, isSelfConnected, setActivationFunction, setDescription, setID, setName, setNetwork, setNeuronCount, setThreshold, setThreshold, setX, setY, toString

Methods inherited from class java.lang.Object

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

Constructor Detail

ContextLayer

public ContextLayer()

Default constructor, mainly so the workbench can easily create a default layer.

ContextLayer

public ContextLayer(ActivationFunction thresholdFunction,
                    boolean hasThreshold,
                    int neuronCount)

Construct a context layer with the parameters specified.

Parameters:

thresholdFunction - The threshold function to use.

hasThreshold - Does this layer have thresholds?

neuronCount - The neuron count to use.

ContextLayer

public ContextLayer(int neuronCount)

Construct a default context layer that has the TANH activation function and the specified number of neurons. Use threshold values.

Parameters:

neuronCount - The number of neurons on this layer.

Method Detail

createPersistor

public Persistor createPersistor()

Create a persistor for this layer.

Specified by:

createPersistor in interface EncogPersistedObject

Overrides:

createPersistor in class BasicLayer

Returns:

The new persistor.

getContext

public NeuralData getContext()

Returns:

The context, or memory of this layer. These will be the values that were just output.

process

public void process(NeuralData pattern)

Called to process input from the previous layer. Simply store the output in the context.

Specified by:

process in interface Layer

Overrides:

process in class BasicLayer

Parameters:

pattern - The pattern to store in the context.

recur

public NeuralData recur()

Called to get the output from this layer when called in a recurrent manor. Simply return the context that was kept from the last iteration.

Specified by:

recur in interface Layer

Overrides:

recur in class BasicLayer

Returns:

The recurrent output.

clearContext

public void clearContext()

Reset the context values back to zero.

Specified by:

clearContext in interface ContextClearable