SciLib.Neuro
Class Nnet

java.lang.Object
  SciLib.Neuro.Nnet

Direct Known Subclasses:

Adaline, Madaline, RbfNet

public class Nnet
extends java.lang.Object

This class defines the base class for all Neural Networks.

Field Summary

protected double	averageError
protected matrix	DesiredOutputMatrix
protected double	errorThreshold
protected Activation	g
protected java.util.Vector<Layer>	hidden
protected integerVector	hiddenSize
protected int	inSize
protected double	learningRate
protected double	momentum
protected int	numberOfHiddenLayers
protected int	NumberOfTestPatterns
protected int	NumberOfTrainingPatterns
protected Layer	output
protected int	outSize
protected matrix	testingData
(package private) vector	testingDesiredOutput
protected matrix	trainingData
(package private) vector	trainingDesiredOutput

Constructor Summary

Nnet()
Make an empty neural network

Nnet(int inS, int outS, int noOfH)
Make a neural network with a specific number of input, output and hidden size

Nnet(int inS, int outS, int noOfH, integerVector hiddenS)
Make a neural network with a specific number of input, output and hidden size

Nnet(int inS, int outS, int noOfH, integerVector hiddenS, Activation act)
Make a neural network with a specific number of input, output, hidden layer and activation function

Method Summary

void	generate() Initialize the parameters of the neural network
double	getErrorThreshold() get error Threshold
java.util.Vector	getHiddenLayers() get hidden layers
integerVector	gethiddenSize() get sizes of the hidden layers
int	getInputSize() get input size
double	getLearningRate() get learning rate
double	getMomentumPar() get momentum
int	getnumberOfHiddenLayers() get number of hidden layers
int	getNumberOfTestPatterns() get number of testing patterns
int	getNumberOfTrainingPatterns() get number of training patterns
Layer	getOutputLayer() get the output layer
int	getOutputSize() get output size
matrix	getTrainingData() get Training Data
vector	getTrainingDesiredOutput() get Training Desired Output
void	makeDesiredOutputMatrix() make a Desired Output matrix handles the case of multiple output
void	makeDesiredOutputMatrix(int type) make a Desired Output matrix handles the case of multiple output
void	makeTrainingSet(int n, int m)
void	readTestingData(java.lang.String filename) read Testing Data file File format Number of hidden layers Size of the input vector Number of the hidden nodes Size of the output vector Number of Training Patterns: No_of_Patterns Data : Training pattern vector : Training_Patterns[k][0:Size_in-1] Desired output (classes): Training_Patterns[k][Size_in]
void	readTrainingData(java.lang.String filename) read Training data File format Example: 3 8 6 4 4 2 256 2.1 3.2 1 0.6 3 -1 2 1.2 1 Number of hidden layers : 3 Size of the input vector : 8 Number of the hidden nodes : 6 4 4 Size of the output vector : 2 Number of Training Patterns: 256 Data : Training pattern vector : Training_Patterns[k][0:Size_in-1] Desired output (classes): TrainingDesiredOutput[k]
void	readWeight(java.lang.String filename) // File format // Number of hidden layers : No_of_hidden_layers // Size of the input vector : Size_in // Number of the hidden nodes : Size_hidden[i] // Size of the output vector : Size_out // Number of Training Patterns: No_of_Patterns // Data : // Hidden Layer Weight : hidden[k]->weigth // Output Layer Weight : output->weight //
void	setActivation(Activation act) set Activation function
void	setErrorThreshold(double epsi) set error threshold
void	setInputSize(int p) set the input size
void	setLearningRate(double eta) set learning rate
void	setMomentumPar(double alpha) set momentum
void	setNumberOfTestPatterns(int n) set number of testing patterns
void	setNumberOfTrainingPatterns(int n) set number of training patterns
void	setOutputSize(int q) set the output size
protected void	setSize(int inS, int outS, int noOfH, integerVector hS) set size to neural network
void	writeWeight(java.lang.String filename) write weight matrix to a file

Methods inherited from class java.lang.Object

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

Field Detail

inSize

protected int inSize

outSize

protected int outSize

numberOfHiddenLayers

protected int numberOfHiddenLayers

hiddenSize

protected integerVector hiddenSize

learningRate

protected double learningRate

momentum

protected double momentum

errorThreshold

protected double errorThreshold

averageError

protected double averageError

NumberOfTestPatterns

protected int NumberOfTestPatterns

NumberOfTrainingPatterns

protected int NumberOfTrainingPatterns

trainingData

protected matrix trainingData

testingData

protected matrix testingData

DesiredOutputMatrix

protected matrix DesiredOutputMatrix

trainingDesiredOutput

vector trainingDesiredOutput

testingDesiredOutput

vector testingDesiredOutput

output

protected Layer output

hidden

protected java.util.Vector<Layer> hidden

g

protected Activation g

Constructor Detail

Nnet

public Nnet()

Make an empty neural network

Nnet

public Nnet(int inS,
            int outS,
            int noOfH)

Make a neural network with a specific number of input, output and hidden size

Parameters:

inS - An integer value, the size of the input vector

outS - An integer value, the size of the output vector

noOfH - An integer value, the number of hidden layers

Nnet

public Nnet(int inS,
            int outS,
            int noOfH,
            integerVector hiddenS)

Make a neural network with a specific number of input, output and hidden size

Parameters:

inS - An integer value, the size of the input vector

outS - An integer value, the size of the output vector

noOfH - An integer value, the number of hidden layers

hiddenS - An integer vector contains the size of the hidden layers

Nnet

public Nnet(int inS,
            int outS,
            int noOfH,
            integerVector hiddenS,
            Activation act)

Make a neural network with a specific number of input, output, hidden layer and activation function

Parameters:

inS - An integer value, the size of the input vector

outS - An integer value, the size of the output vector

noOfH - An integer value, the number of hidden layers

hiddenS - An integer vector contains the size of the hidden layers

act - An object that implements the Activation interface.

Method Detail

setSize

protected void setSize(int inS,
                       int outS,
                       int noOfH,
                       integerVector hS)

set size to neural network

Parameters:

inS - An integer value, the size of the input vector

outS - An integer value, the size of the output vector

noOfH - An integer value, the number of hidden layers

hS - An integer vector contains the size of the hidden layers

generate

public void generate()

Initialize the parameters of the neural network

getLearningRate

public double getLearningRate()

get learning rate

Returns:

a double value

getErrorThreshold

public double getErrorThreshold()

get error Threshold

Returns:

a double value

getMomentumPar

public double getMomentumPar()

get momentum

Returns:

a double value

getNumberOfTrainingPatterns

public int getNumberOfTrainingPatterns()

get number of training patterns

Returns:

an integer value

getNumberOfTestPatterns

public int getNumberOfTestPatterns()

get number of testing patterns

Returns:

an integer value

getnumberOfHiddenLayers

public int getnumberOfHiddenLayers()

get number of hidden layers

Returns:

an integer value

getInputSize

public int getInputSize()

get input size

Returns:

an integer value

getOutputSize

public int getOutputSize()

get output size

Returns:

an integer value

gethiddenSize

public integerVector gethiddenSize()

get sizes of the hidden layers

Returns:

an integer vector

getOutputLayer

public Layer getOutputLayer()

get the output layer

Returns:

a Layer

getHiddenLayers

public java.util.Vector getHiddenLayers()

get hidden layers

Returns:

a Vector of hidden layers

getTrainingDesiredOutput

public vector getTrainingDesiredOutput()

get Training Desired Output

Returns:

a vector

getTrainingData

public matrix getTrainingData()

get Training Data

Returns:

a matrix

setLearningRate

public void setLearningRate(double eta)

set learning rate

Parameters:

eta - double value

setMomentumPar

public void setMomentumPar(double alpha)

set momentum

Parameters:

alpha - double value

setNumberOfTrainingPatterns

public void setNumberOfTrainingPatterns(int n)

set number of training patterns

Parameters:

n - an integer value

setNumberOfTestPatterns

public void setNumberOfTestPatterns(int n)

set number of testing patterns

Parameters:

n - an integer value

setErrorThreshold

public void setErrorThreshold(double epsi)

set error threshold

Parameters:

epsi - double value

setInputSize

public void setInputSize(int p)

set the input size

Parameters:

p - an integer value

setOutputSize

public void setOutputSize(int q)

set the output size

Parameters:

q - an integer value

setActivation

public void setActivation(Activation act)

set Activation function

Parameters:

act - an Activation function object

readTrainingData

public void readTrainingData(java.lang.String filename)

read Training data File format Example: 3 8 6 4 4 2 256 2.1 3.2 1 0.6 3 -1 2 1.2 1 Number of hidden layers : 3 Size of the input vector : 8 Number of the hidden nodes : 6 4 4 Size of the output vector : 2 Number of Training Patterns: 256 Data : Training pattern vector : Training_Patterns[k][0:Size_in-1] Desired output (classes): TrainingDesiredOutput[k]

readTestingData

public void readTestingData(java.lang.String filename)

read Testing Data file File format Number of hidden layers Size of the input vector Number of the hidden nodes Size of the output vector Number of Training Patterns: No_of_Patterns Data : Training pattern vector : Training_Patterns[k][0:Size_in-1] Desired output (classes): Training_Patterns[k][Size_in]

readWeight

public void readWeight(java.lang.String filename)

// File format // Number of hidden layers : No_of_hidden_layers // Size of the input vector : Size_in // Number of the hidden nodes : Size_hidden[i] // Size of the output vector : Size_out // Number of Training Patterns: No_of_Patterns // Data : // Hidden Layer Weight : hidden[k]->weigth // Output Layer Weight : output->weight //

writeWeight

public void writeWeight(java.lang.String filename)

write weight matrix to a file

makeDesiredOutputMatrix

public void makeDesiredOutputMatrix()

make a Desired Output matrix handles the case of multiple output

makeDesiredOutputMatrix

public void makeDesiredOutputMatrix(int type)

make a Desired Output matrix handles the case of multiple output

makeTrainingSet

public void makeTrainingSet(int n,
                            int m)