com.huguesjohnson.jail.neuralNetwork
Class BackPropagationNetwork

java.lang.Object
  |
  +--com.huguesjohnson.jail.neuralNetwork.NeuralNetwork
        |
        +--com.huguesjohnson.jail.neuralNetwork.BackPropagationNetwork

All Implemented Interfaces:

java.lang.Runnable, java.io.Serializable

public class BackPropagationNetwork

extends NeuralNetwork

BackPropagationNetwork

See Also:

Serialized Form

Constructor Summary

BackPropagationNetwork(int inputLayerSize, int hiddenLayerSize, int outputLayerSize, int inputSize, double punishRate, double rewardRate, double errorThreshold, double minWeight, double maxWeight)
constructor

Method Summary

double[]	getDesiredOutput() returns the desired output of the network
double	getError() returns the error of the network
com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer	getHiddenLayer() returns the hidden layer
com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer	getInputLayer() returns the input layer
double[]	getOutput() returns the output of the network
com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer	getOutputLayer() returns the output layer
void	networkRunComplete() method invoked when the network run is complete
void	run() run the network
void	setDesiredOutput(double[] desiredOutput) sets the desired output of the network
void	setHiddenLayer(com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer hiddenLayer) sets the hidden layer, automatically adjusts links between input layer and hidden layer and hidden layer and output layer
void	setInput(double[] input) sets the input of the network
void	setInputLayer(com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer inputLayer) sets the input layer, automatically adjusts links between input layer and hidden layer
void	setOutputLayer(com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer outputLayer) sets the output layer, automatically adjusts links between hidden layer and output layer
void	train(int iterations) train the network

Methods inherited from class com.huguesjohnson.jail.neuralNetwork.NeuralNetwork

finalize, log, start, stop, toString

Methods inherited from class java.lang.Object

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

Constructor Detail

BackPropagationNetwork

public BackPropagationNetwork(int inputLayerSize,
                              int hiddenLayerSize,
                              int outputLayerSize,
                              int inputSize,
                              double punishRate,
                              double rewardRate,
                              double errorThreshold,
                              double minWeight,
                              double maxWeight)

constructor

Parameters:

inputLayerSize - the number of neurons in the input layer

hiddenLayerSize - the number of neurons in the hidden layer

outputLayerSize - the number of neurons in the output layer

inputSize - the number of inputs to the layer

punishRate - the punish rate for every neuron in the layer

rewardRate - the reward rate for every neuron in the layer

errorThreshold - the error threshold for every neuron in the layer

minWeight - minimum weight on a neuron link

maxWeight - maximum weight on a neuron link

Method Detail

getOutput

public double[] getOutput()

returns the output of the network

Specified by:

getOutput in class NeuralNetwork

Returns:

output of the network

getError

public double getError()

returns the error of the network

Specified by:

getError in class NeuralNetwork

Returns:

error of the network

setDesiredOutput

public void setDesiredOutput(double[] desiredOutput)

sets the desired output of the network

Specified by:

setDesiredOutput in class NeuralNetwork

Parameters:

desiredOutput - the desired output of the network

setInput

public void setInput(double[] input)

sets the input of the network

Specified by:

setInput in class NeuralNetwork

Parameters:

input - the input of the network

getInputLayer

public final com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer getInputLayer()

returns the input layer

Returns:

input layer

getHiddenLayer

public final com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer getHiddenLayer()

returns the hidden layer

Returns:

hidden layer

getOutputLayer

public final com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer getOutputLayer()

returns the output layer

Returns:

output layer

setInputLayer

public final void setInputLayer(com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer inputLayer)

sets the input layer, automatically adjusts links between input layer and hidden layer

Parameters:

inputLayer - new input layer

setHiddenLayer

public final void setHiddenLayer(com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer hiddenLayer)

sets the hidden layer, automatically adjusts links between input layer and hidden layer and hidden layer and output layer

Parameters:

hiddenLayer - new hidden layer

setOutputLayer

public final void setOutputLayer(com.huguesjohnson.jail.neuralNetwork.BackPropagationLayer outputLayer)

sets the output layer, automatically adjusts links between hidden layer and output layer

Parameters:

outputLayer - new output layer

train

public void train(int iterations)

train the network

Specified by:

train in class NeuralNetwork

Parameters:

iterations - number of training iterations to run

run

public void run()

run the network

Specified by:

run in interface java.lang.Runnable

Overrides:

run in class NeuralNetwork

networkRunComplete

public void networkRunComplete()

method invoked when the network run is complete

getDesiredOutput

public double[] getDesiredOutput()

returns the desired output of the network

Returns:

desired output of the network