keel.Algorithms.Neural_Networks.gmdh

Class Network

java.lang.Object

keel.Algorithms.Neural_Networks.gmdh.Network

public class Network
extends java.lang.Object

Class representing a Nural Network

Since:

JDK1.5

Version:

0.1

Author:

Written by Nicolas Garcia Pedrajas (University of Cordoba) 27/02/2007

Field Summary

Fields

Modifier and Type	Field and Description
double	a Transfer function parameters
double[][]	activation Output of each node
double	b_htan Transfer function parameters
double	b_log Transfer function parameters
double[][]	delta Delta weights
double[][][]	momentum Momentum term
int[]	Nhidden Number of units in each layer
int	Ninputs Number of units in each layer
int	Nlayers Number of layers
int	Noutputs Number of units in each layer
java.lang.String[]	transfer Transfer function of each layer (LOG | HTAN | LINEAR)
double[][][]	w Matrix of weights

Constructor Summary

Constructors

Constructor and Description
Network() Empty Constructor.
Network(Parameters global) Constructor that receives parameters

Method Summary

Modifier and Type	Method and Description
void	GenerateOutput(double[] input) Generate output at activation using an input
void	GenerateOutput(double[] input, double[] output) Generate output using an input
void	LoadNetwork(java.lang.String file_name) Load network weights from file
boolean	NetClassifyPattern(double[] pattern) Check if a pattern is correctly classified
int	NetGetClassOfPattern(double[] pattern) Return the class where a pattern is classified
void	PrintWeights() Print weights to screen.
void	SaveNetwork(java.lang.String file_name, boolean append) Save network weights
void	SaveOutputFile(java.lang.String file_name, double[][] data, int n, java.lang.String problem) Save output data to file
double	TestNetworkInClassification(Parameters global, double[][] data, int npatterns) Test network in classification
double	TestNetworkInRegression(Parameters global, double[][] data, int npatterns) Test network in regression.
void	TrainNetwork(Parameters global, double[][] data, int npatterns) Train network without cross validation
void	TrainNetworkWithCrossvalidation(Parameters global, Data data) Train the Network using cross validation

Methods inherited from class java.lang.Object

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

Field Detail

Nlayers

public int Nlayers

Number of layers

Ninputs

public int Ninputs

Number of units in each layer

Noutputs

public int Noutputs

Number of units in each layer

Nhidden

public int[] Nhidden

Number of units in each layer

w

public double[][][] w

Matrix of weights

momentum

public double[][][] momentum

Momentum term

delta

public double[][] delta

Delta weights

activation

public double[][] activation

Output of each node

a

public final double a

Transfer function parameters

See Also:

Constant Field Values

b_log

public final double b_log

Transfer function parameters

See Also:

Constant Field Values

b_htan

public final double b_htan

Transfer function parameters

See Also:

Constant Field Values

transfer

public java.lang.String[] transfer

Transfer function of each layer (LOG | HTAN | LINEAR)

Constructor Detail

Network

public Network()

Empty Constructor.

Network

public Network(Parameters global)

Constructor that receives parameters

Parameters:

global - Global Definition parameters

Method Detail

TrainNetworkWithCrossvalidation

public void TrainNetworkWithCrossvalidation(Parameters global,
                                            Data data)

Train the Network using cross validation

Parameters:

global - Global Definition parameters

data - Input data

TrainNetwork

public void TrainNetwork(Parameters global,
                         double[][] data,
                         int npatterns)

Train network without cross validation

Parameters:

global - Global Definition parameters

data - Input data

npatterns - Number of patterns

TestNetworkInClassification

public double TestNetworkInClassification(Parameters global,
                                          double[][] data,
                                          int npatterns)

Test network in classification

Parameters:

global - Global Definition parameters

data - Input data

npatterns - Number of patterns

Returns:

Number of correctly classified per cent

TestNetworkInRegression

public double TestNetworkInRegression(Parameters global,
                                      double[][] data,
                                      int npatterns)

Test network in regression.

Parameters:

global - Global Definition parameters

data - Input data

npatterns - Number of patterns

Returns:

RMS Error

GenerateOutput

public void GenerateOutput(double[] input)

Generate output at activation using an input

Parameters:

input - Input data

GenerateOutput

public void GenerateOutput(double[] input,
                           double[] output)

Generate output using an input

Parameters:

input - Input data

output - Output data

SaveNetwork

public void SaveNetwork(java.lang.String file_name,
                        boolean append)

Save network weights

Parameters:

file_name - Output file name

append - Append or overwrite flag

LoadNetwork

public void LoadNetwork(java.lang.String file_name)

Load network weights from file

Parameters:

file_name - Input file name

PrintWeights

public void PrintWeights()

Print weights to screen. Not used

NetClassifyPattern

public boolean NetClassifyPattern(double[] pattern)

Check if a pattern is correctly classified

Parameters:

pattern - Pattern to check

Returns:

true if it is correctly classified. False otherwise

NetGetClassOfPattern

public int NetGetClassOfPattern(double[] pattern)

Return the class where a pattern is classified

Parameters:

pattern - Pattern to check

Returns:

Class index

SaveOutputFile

public void SaveOutputFile(java.lang.String file_name,
                           double[][] data,
                           int n,
                           java.lang.String problem)

Save output data to file

Parameters:

file_name - Output file name

data - Input data

n - Data matrix order (number of rows and columns)

problem - Type of problem (CLASSIFICATION | REGRESSION )