jhplot

Class HNeuralNet

java.lang.Object

jhplot.HNeuralNet

public class HNeuralNetextends Object

Neural Netwrork calculations. Based on Backpropagation.

Constructor Summary

Constructors

Constructor and Description
HNeuralNet() Create a network net and set name for the network

Method Summary

Modifier and Type	Method and Description
void	addFeedForwardLayer(int neuronCount) Construct this layer with a sigmoid threshold function.
void	addFeedForwardLayerWithBias(int neuronCount) Construct this layer with a sigmoid threshold function.
AnalyzeNetwork	analyzeNetwork() Analyse the current neural network.
void	doc() Show online documentation.
BasicNetwork	editNetwork() Edit a neural net in a frame
BasicNeuralDataSet	getData() Get data
ArrayList<Double>	getEpochError() Returns errors for each epoch.
BasicNetwork	getNetwork() Return neural net back.
MLData	predict(MLData input) Evaluate data using current NN
P0D	predict(P0D input) Generate prediction for input data
PND	predict(PND input) Generate predictions for all input data.
int	read(String file) Read a neural net from a file.
void	reset() Reset the weight matrix and the thresholds.
String	save(String file) Save current status of neural net.
void	setData(double[][] input) Construct a data set from an input
void	setData(double[][] input, double[][] ideal) Construct a data set from an input and idea array.
void	setData(PND input) Set data
void	setData(PND input, PND ideal) Set data for training.
void	show() Show Net in EncodeDocument.
void	showNetwork() Show a neural net in a frame.
void	showWeights() Show a neural net weights in a separate frame.
PND	standardize(PND input) Standardize each column.
int	trainBackpropagation(boolean isShow, int maxEpoch, double learnRate, double momentum, double errorMinEpoch) Training neural network.Construct a backpropagation trainer.

Methods inherited from class java.lang.Object

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

Constructor Detail

HNeuralNet

public HNeuralNet()

Create a network net and set name for the network

Parameters:

name - name for the network

Method Detail

reset

public void reset()

Reset the weight matrix and the thresholds.

addFeedForwardLayer

public void addFeedForwardLayer(int neuronCount)

Construct this layer with a sigmoid threshold function. Use sigmoid for activation.

Parameters:

neuronCount - How many neurons in this layer

addFeedForwardLayerWithBias

public void addFeedForwardLayerWithBias(int neuronCount)

Construct this layer with a sigmoid threshold function. Use sigmoid for activation.

Parameters:

neuronCount - How many neurons in this layer

setData

public void setData(double[][] input,                    double[][] ideal)

Construct a data set from an input and idea array. Used for supervized training.

Parameters:

input - The input into the neural network for training.

ideal - The ideal output for training.

setData

public void setData(double[][] input)

Construct a data set from an input

Parameters:

input - The input into the neural network for training.

setData

public void setData(PND input,                    PND ideal)

Set data for training.

Parameters:

input - input data set

ideal - expected resul.

setData

public void setData(PND input)

Set data

Parameters:

input - input data set

standardize

public PND standardize(PND input)

Standardize each column. This means S(i)= (X(i) - mean) / std fot each column in PND;

Parameters:

input - PND

Returns:

new PND after standardize

getData

public BasicNeuralDataSet getData()

Get data

Returns:

data

predict

public MLData predict(MLData input)

Evaluate data using current NN

Returns:

data

predict

public P0D predict(P0D input)

Generate prediction for input data

Parameters:

input - input data for predictions

predict

public PND predict(PND input)

Generate predictions for all input data. Assumes that the predicted array has less then 3 dimensions.

Parameters:

input - input data for prediction

Returns:

data with predictions

trainBackpropagation

public int trainBackpropagation(boolean isShow,                                int maxEpoch,                                double learnRate,                                double momentum,                                double errorMinEpoch)

Training neural network.Construct a backpropagation trainer. Typical example: train(5000, 0.1, 0.25, 0.001);

Parameters:

isShow - Show learning on a pop-up plot

maxEpoch - maximum number of epochs

learnRate - The rate at which the weight matrix will be adjusted based on learning.

momentum - The influence that previous iteration's training deltas will have on the current iteration.

errorMinEpoch - min error for epoch.

Returns:

returns the epoch at which training was stopped.

save

public String save(String file)

Save current status of neural net.

Parameters:

file - File name

Returns:

what is done

read

public int read(String file)

Read a neural net from a file.

Parameters:

file - File name

Returns:

0 if it is OK. -1 if file not found; -2: if NN not found.

getNetwork

public BasicNetwork getNetwork()

Return neural net back.

Returns:

network

showNetwork

public void showNetwork()

Show a neural net in a frame.

showWeights

public void showWeights()

Show a neural net weights in a separate frame.

analyzeNetwork

public AnalyzeNetwork analyzeNetwork()

Analyse the current neural network.

Returns:

analyzer

editNetwork

public BasicNetwork editNetwork()

Edit a neural net in a frame

show

public void show()

Show Net in EncodeDocument.

getEpochError

public ArrayList<Double> getEpochError()

Returns errors for each epoch. If the max epoch number was set in the train() method. The array may have less entries if learning has reached the minimum error.

Returns:

arrays of errors for each epoch

doc

public void doc()

Show online documentation.