org.yccheok.numrecognition
Class NumberNeuralNetworkTrainer

java.lang.Object
  org.yccheok.numrecognition.NumberNeuralNetworkTrainer

All Implemented Interfaces:

java.util.EventListener, org.joone.engine.NeuralNetListener

public class NumberNeuralNetworkTrainer
extends java.lang.Object
implements org.joone.engine.NeuralNetListener

This is the neural network used to perform training.

Field Summary

private NumberImageFeatureFactory	featureFactory Feature factory for this neural network.
private org.joone.engine.SigmoidLayer	hidden The hidden layer of neurons.
private org.joone.engine.SigmoidLayer	input The input layer of neurons.
private org.joone.engine.Monitor	monitor Monitor.
private org.joone.engine.SigmoidLayer	output The output layer of neurons.
private org.joone.engine.FullSynapse	t1 FullSynapse.
private org.joone.engine.FullSynapse	t2 Full Synapse.
private org.joone.engine.learning.TeachingSynapse	trainer Used to train the neural network.

Constructor Summary

NumberNeuralNetworkTrainer(NumberImageFeatureFactory featureFactory, NeuralNetworkParameter neuralNetworkParameter)
Creates a new instance of NumberNeuralNetworkTrainer with specified feature factory and feature parameter.

Method Summary

void	cicleTerminated(org.joone.engine.NeuralNetEvent e) Please refer to JOONE doc.
void	errorChanged(org.joone.engine.NeuralNetEvent e) Please refer to JOONE doc.
int	GenerateTraingDataFile(NumberCategory[] numberCategory, java.lang.String trainingDataFileName) Generate the training data file.
int	getHiddenRows() Returns number of hidden neurons.
void	netStarted(org.joone.engine.NeuralNetEvent e) Please refer to JOONE doc.
void	netStopped(org.joone.engine.NeuralNetEvent e) Please refer to JOONE doc.
void	netStoppedError(org.joone.engine.NeuralNetEvent e, java.lang.String error) Please refer to JOONE doc.
boolean	saveNeuralNetwork(java.lang.String fileName) Save the neural network.
void	train(NumberCategory[] numberCategory, java.lang.String errorOutputFile, int totCicles) Train the neural network.
void	train(java.lang.String inputFile, java.lang.String errorOutputFile, int numOfInputRow, int numOfPattern, int totCicles) Train the neural network.

Methods inherited from class java.lang.Object

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

Field Detail

input

private org.joone.engine.SigmoidLayer input

The input layer of neurons.

hidden

private org.joone.engine.SigmoidLayer hidden

The hidden layer of neurons.

output

private org.joone.engine.SigmoidLayer output

The output layer of neurons.

trainer

private org.joone.engine.learning.TeachingSynapse trainer

Used to train the neural network.

t1

private org.joone.engine.FullSynapse t1

FullSynapse.

t2

private org.joone.engine.FullSynapse t2

Full Synapse.

monitor

private org.joone.engine.Monitor monitor

Monitor.

featureFactory

private NumberImageFeatureFactory featureFactory

Feature factory for this neural network.

Constructor Detail

NumberNeuralNetworkTrainer

public NumberNeuralNetworkTrainer(NumberImageFeatureFactory featureFactory,
                                  NeuralNetworkParameter neuralNetworkParameter)

Creates a new instance of NumberNeuralNetworkTrainer with specified feature factory and feature parameter.

Parameters:

featureFactory - Feature factory.

neuralNetworkParameter - Neural network parameter.

Method Detail

train

public void train(NumberCategory[] numberCategory,
                  java.lang.String errorOutputFile,
                  int totCicles)

Train the neural network.

Parameters:

numberCategory - Array of number category.

errorOutputFile - File to be written to for the error rate information during training.

totCicles - Total iterations during training.

GenerateTraingDataFile

public int GenerateTraingDataFile(NumberCategory[] numberCategory,
                                  java.lang.String trainingDataFileName)

Generate the training data file.

Parameters:

numberCategory - Array of number category.

trainingDataFileName - File name of the training file to be generated.

Returns:

true if the training file can be generated successfully.

train

public void train(java.lang.String inputFile,
                  java.lang.String errorOutputFile,
                  int numOfInputRow,
                  int numOfPattern,
                  int totCicles)

Train the neural network.

Parameters:

inputFile - Training data file name.

errorOutputFile - File to be written to for the error rate information during training.

numOfInputRow - Size per training pattern. This will determine the size of input neurons.

numOfPattern - Number of patterns. This is the number of lines of the training data file.

totCicles - Total iterations during training.

netStopped

public void netStopped(org.joone.engine.NeuralNetEvent e)

Please refer to JOONE doc.

Specified by:

netStopped in interface org.joone.engine.NeuralNetListener

Parameters:

e - Please refer to JOONE doc.

cicleTerminated

public void cicleTerminated(org.joone.engine.NeuralNetEvent e)

Please refer to JOONE doc.

Specified by:

cicleTerminated in interface org.joone.engine.NeuralNetListener

Parameters:

e - Please refer to JOONE doc.

netStarted

public void netStarted(org.joone.engine.NeuralNetEvent e)

Please refer to JOONE doc.

Specified by:

netStarted in interface org.joone.engine.NeuralNetListener

Parameters:

e - Please refer to JOONE doc.

errorChanged

public void errorChanged(org.joone.engine.NeuralNetEvent e)

Please refer to JOONE doc.

Specified by:

errorChanged in interface org.joone.engine.NeuralNetListener

Parameters:

e - Please refer to JOONE doc.

netStoppedError

public void netStoppedError(org.joone.engine.NeuralNetEvent e,
                            java.lang.String error)

Please refer to JOONE doc.

Specified by:

netStoppedError in interface org.joone.engine.NeuralNetListener

Parameters:

e - Please refer to JOONE documentation.

error - Please refer to JOONE documentation.

saveNeuralNetwork

public boolean saveNeuralNetwork(java.lang.String fileName)

Save the neural network.

Parameters:

fileName - File name to be saved into.

Returns:

Whether the neural network can be saved successfully.

getHiddenRows

public int getHiddenRows()

Returns number of hidden neurons.

Returns:

Number of hidden neurons.