Class RNN

java.lang.Object
  Network
      RNN

public class RNN

extends Network

The RNN class is based on the Elman Network, which has recurrency at the Hidden Layer. In addition, there is a self loop at the context layer

Author:

Raymond McBride

Constructor Summary

RNN(int inputs, int hiddens, double memoryDepth, int slope, double learningRate, double momentum, int totalEpochs, String fileID)
This constructor for the RNN specifies the number of Input Neurons and Hidden Neurons, the memory depth, the slope of their activation functions, the learning rate, the momentum, the number of training, epochs and set a network topology id for use with log files.

Method Summary

protected void	adjustHiddenWeights() Adjusts the weights of the Synapses between the Hidden and Input Neurons and Bias Neurons
protected void	adjustOutputWeights() Adjusts the weights of the Synapses between the Output and Hidden Neurons and Bias Neurons
protected double	calculateError() Calculates the output error
protected void	calculateHiddenError() Calculates the hidden error term
protected void	calculateHiddenOutput() Calculates the output of the HiddenNeurons
protected void	calculateHiddenWeightChange() Calculates the weight change to be made to the Synapses between the Hidden and Input Neurons and Bias Neurons
protected void	calculateOutputError() Calculates the output error term
protected void	calculateOutputWeightChange() Calculates the weight change to be made to the Synapses between the Output and Hidden Neurons and Bias Neurons
protected void	connectNeurons() Connects the Neurons
protected void	createNeurons() Creates the Neurons
protected void	createNeurons(double memoryDepth) Creates the Neurons
protected BiasNeuron	getBiasHidden() Gets the BiasNeuron to the HiddenNeurons
protected BiasNeuron	getBiasOutput() Gets the BiasNeuron to the OutputNeuron
protected Synapse[]	getBiasToHidden() Gets the Synapses between the BiasNeuron and the HiddenNeurons
protected Synapse	getBiasToHidden(int i) Gets a Synapse between the BiasNeuron and a HiddenNeuron
protected Synapse	getBiasToOutput() Gets the Synapse connecting the BiasNeuron to the OutputNeuron
protected double	getHiddenErrorTerm(int i) Gets a hiddenErrorTerm
protected Neuron	getHiddenNeuron(int i) Gets a HiddenNeuron
protected Neuron[]	getHiddenNeurons() Gets the HiddenNeurons
protected Synapse[]	getHiddenToOutput() Gets the Synapses between the HiddenNeurons and the OutputNeuron
protected Synapse	getHiddenToOutput(int i) Gets a Synapse between a HiddenNeuron and the OutputNeuron
protected double[]	getInputData() Gets the input data
protected double	getInputData(int i) Gets input data at a given position
protected Neuron	getInputNeuron(int i) Gets an InputNeuron
protected Neuron[]	getInputNeurons() Gets the InputNeurons
protected Synapse[][]	getInputToHidden() Gets all the Synapses between the InputNeurons and the HiddenNeurons
protected Synapse[]	getInputToHidden(int i) Gets the Synapses between an InputNeuron and the HiddenNeurons
protected Synapse	getInputToHidden(int i, int j) Gets the Synapse between an InputNeuron and a HiddenNeuron
protected double	getLearningRate() Gets the learning rate
protected double	getMomentum() Gets the momentum
protected int	getNextInput() Gets the next input position
protected double	getOutputErrorTerm() Gets the OutputErrorTerm
protected int	getSlope() Gets the slope
protected void	initialise() Initialises the network with data
protected void	sendToHidden() Transfers the weighted values to the Hidden Neurons
protected void	sendToOutput() Transfers the weighted values to the Output Neuron
protected void	setBiasHidden(BiasNeuron biasHidden) Sets the BiasNeuron to the HiddenNeurons
protected void	setBiasOutput(BiasNeuron biasOutput) Sets the BiasNeuron to the OutputNeuron
protected void	setHiddenErrorTerm(double errorTerm, int i) Sets a hiddenErrorTerm
protected void	setHiddenNeuron(HiddenNeuron hiddenNeuron, int i) Sets a HiddenNeuron
protected void	setInputNeuron(InputNeuron inputNeuron, int i) Sets an InputNeuron
protected void	setOutputNeuron(OutputNeuron outputNeuron) Sets the OutputNeuron
protected void	setTargetOutput(double targetOutput) Sets the target output
void	test(double[] data) Starts testing the network
void	train(double[] data) Trains the network
void	validate(double[] data) Starts validating the network

Methods inherited from class java.lang.Object

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

Constructor Detail

RNN

public RNN(int inputs,
           int hiddens,
           double memoryDepth,
           int slope,
           double learningRate,
           double momentum,
           int totalEpochs,
           String fileID)

This constructor for the RNN specifies the number of Input Neurons and Hidden Neurons, the memory depth, the slope of their activation functions, the learning rate, the momentum, the number of training, epochs and set a network topology id for use with log files.

Parameters:

inputs - The number of Input Neurons

hiddens - The number of Hidden Neurons

memoryDepth - The memory depth

slope - The slope of their activation functions

learningRate - The learning rate

momentum - The momentum

totalEpochs - The number of training epochs

fileID - The network topology id

Method Detail

createNeurons

protected void createNeurons(double memoryDepth)

Creates the Neurons

Parameters:

memoryDepth - Their memory depth

connectNeurons

protected void connectNeurons()

Connects the Neurons

sendToHidden

protected void sendToHidden()

Transfers the weighted values to the Hidden Neurons

sendToOutput

protected void sendToOutput()

Transfers the weighted values to the Output Neuron

initialise

protected void initialise()

Initialises the network with data

createNeurons

protected void createNeurons()

Creates the Neurons

calculateOutputError

protected void calculateOutputError()

Calculates the output error term

calculateHiddenError

protected void calculateHiddenError()

Calculates the hidden error term

calculateOutputWeightChange

protected void calculateOutputWeightChange()

Calculates the weight change to be made to the Synapses between the Output and Hidden Neurons and Bias Neurons

adjustOutputWeights

protected void adjustOutputWeights()

Adjusts the weights of the Synapses between the Output and Hidden Neurons and Bias Neurons

calculateHiddenWeightChange

protected void calculateHiddenWeightChange()

Calculates the weight change to be made to the Synapses between the Hidden and Input Neurons and Bias Neurons

adjustHiddenWeights

protected void adjustHiddenWeights()

Adjusts the weights of the Synapses between the Hidden and Input Neurons and Bias Neurons

calculateHiddenOutput

protected void calculateHiddenOutput()

Calculates the output of the HiddenNeurons

calculateError

protected double calculateError()

Calculates the output error

train

public void train(double[] data)

Trains the network

test

public void test(double[] data)

Starts testing the network

validate

public void validate(double[] data)

Starts validating the network

getSlope

protected int getSlope()

Gets the slope

Returns:

slope

getLearningRate

protected double getLearningRate()

Gets the learning rate

Returns:

learningRate

getMomentum

protected double getMomentum()

Gets the momentum

Returns:

momentum

setTargetOutput

protected void setTargetOutput(double targetOutput)

Sets the target output

Parameters:

targetOutput - the target output

setOutputNeuron

protected void setOutputNeuron(OutputNeuron outputNeuron)

Sets the OutputNeuron

Parameters:

outputNeuron - the OutputNeuron

getOutputErrorTerm

protected double getOutputErrorTerm()

Gets the OutputErrorTerm

Returns:

outputErrorTerm

getHiddenErrorTerm

protected double getHiddenErrorTerm(int i)

Gets a hiddenErrorTerm

Parameters:

i - its position

Returns:

hiddenErrorTerm an array of error terms

setHiddenErrorTerm

protected void setHiddenErrorTerm(double errorTerm,
                                  int i)

Sets a hiddenErrorTerm

Parameters:

errorTerm - the new error term

i - its position

getNextInput

protected int getNextInput()

Gets the next input position

Returns:

nextInput

getBiasToOutput

protected Synapse getBiasToOutput()

Gets the Synapse connecting the BiasNeuron to the OutputNeuron

Returns:

biasToOutput

setBiasHidden

protected void setBiasHidden(BiasNeuron biasHidden)

Sets the BiasNeuron to the HiddenNeurons

Parameters:

biasHidden - the BiasNeuron to the HiddenNeurons

getBiasHidden

protected BiasNeuron getBiasHidden()

Gets the BiasNeuron to the HiddenNeurons

Returns:

biasHidden the BiasNeuron to the HiddenNeurons

setBiasOutput

protected void setBiasOutput(BiasNeuron biasOutput)

Sets the BiasNeuron to the OutputNeuron

Parameters:

biasOutput - the BiasNeuron to the OutputNeuron

getBiasOutput

protected BiasNeuron getBiasOutput()

Gets the BiasNeuron to the OutputNeuron

Returns:

biasOutput the BiasNeuron to the OutputNeuron

getBiasToHidden

protected Synapse[] getBiasToHidden()

Gets the Synapses between the BiasNeuron and the HiddenNeurons

Returns:

biasToHidden the array of Synapses

getBiasToHidden

protected Synapse getBiasToHidden(int i)

Gets a Synapse between the BiasNeuron and a HiddenNeuron

Parameters:

i - its position

Returns:

the Synapse at that position

getHiddenToOutput

protected Synapse getHiddenToOutput(int i)

Gets a Synapse between a HiddenNeuron and the OutputNeuron

Parameters:

i - its position

Returns:

the Synapse at the position

getHiddenToOutput

protected Synapse[] getHiddenToOutput()

Gets the Synapses between the HiddenNeurons and the OutputNeuron

Returns:

hiddenToOutput the array of Synapses

getInputToHidden

protected Synapse[][] getInputToHidden()

Gets all the Synapses between the InputNeurons and the HiddenNeurons

Returns:

inputToHidden the 2D array of Synapses

getInputToHidden

protected Synapse[] getInputToHidden(int i)

Gets the Synapses between an InputNeuron and the HiddenNeurons

Parameters:

i - its position

Returns:

the array of Synapses at that position

getInputToHidden

protected Synapse getInputToHidden(int i,
                                   int j)

Gets the Synapse between an InputNeuron and a HiddenNeuron

Parameters:

i - its x position

j - its y position

Returns:

the Synapse at that position

getInputNeuron

protected Neuron getInputNeuron(int i)

Gets an InputNeuron

Parameters:

i - its position

Returns:

the InputNeuron

getInputNeurons

protected Neuron[] getInputNeurons()

Gets the InputNeurons

Returns:

inputNeurons the array of InputNeurons

setInputNeuron

protected void setInputNeuron(InputNeuron inputNeuron,
                              int i)

Sets an InputNeuron

Parameters:

inputNeuron - the new InputNeuron

i - its position

getHiddenNeuron

protected Neuron getHiddenNeuron(int i)

Gets a HiddenNeuron

Parameters:

i - its position

Returns:

the HiddenNeuron

getHiddenNeurons

protected Neuron[] getHiddenNeurons()

Gets the HiddenNeurons

Returns:

hiddenNeurons the array of HiddenNeurons

setHiddenNeuron

protected void setHiddenNeuron(HiddenNeuron hiddenNeuron,
                               int i)

Sets a HiddenNeuron

Parameters:

hiddenNeuron - the new HiddenNeuron

i - its position

getInputData

protected double getInputData(int i)

Gets input data at a given position

Parameters:

i - its position

Returns:

the double at that position

getInputData

protected double[] getInputData()

Gets the input data

Returns:

inputData