keel.Algorithms.Shared.ClassicalOptim

Class GCNet

java.lang.Object

keel.Algorithms.Shared.ClassicalOptim.GCNet

public class GCNet
extends java.lang.Object

 Wrapper for a perceptron (ConjGradNN). Also this class allows to call the desired training method of the aggregated percetron:
  * nntrain: for invoking Conjugated Gradient.
  * nntrainDG: for invoking the Descendent Gradient.
 

Since:

JDK1.4

Version:

1.0

Author:

Written by Luciano Sánchez (University of Oviedo) 27/02/2004, Modified by Enrique A. de la Cal (University of Oviedo) 13/12/2008

Constructor Summary

Constructors

Constructor and Description
GCNet()

Method Summary

Modifier and Type	Method and Description
double[]	nnoutput(double[] x) Calculated the output of present perceptron with input x and returns it in original scale.
double	nntrain(int nInputs, int nOutputs, double[][] examples, double[][] outputs, int[] topology, double[] weights, Randomize r) trains a perceptron with Conjugated Gradient algorithm and returns the mean square error of neural network output compared to expected output.
double	nntrainGD(int nInputs, int nOutputs, double[][] examples, double[][] outputs, int[] topology, double[] weights, Randomize r) trains a perceptron with Conjugated Descendent algorithm and returns the mean square error of neural network output compared to expected output.
java.lang.String	printNet() Returns the weights as a String for printing or storing purposes

Methods inherited from class java.lang.Object

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

Constructor Detail

GCNet

public GCNet()

Method Detail

nntrain

public double nntrain(int nInputs,
                      int nOutputs,
                      double[][] examples,
                      double[][] outputs,
                      int[] topology,
                      double[] weights,
                      Randomize r)

trains a perceptron with Conjugated Gradient algorithm and returns the mean square error of neural network output compared to expected output.

Parameters:

nInputs - number of inputs in first layer

nOutputs - number of output in output layer

examples - training examples

outputs - expected outputs

topology - net topology (cardinality of the hidder layers)

weights - net weights

r - random generator

Returns:

the mean square error of neural network output compared to expected output.

nntrainGD

public double nntrainGD(int nInputs,
                        int nOutputs,
                        double[][] examples,
                        double[][] outputs,
                        int[] topology,
                        double[] weights,
                        Randomize r)

trains a perceptron with Conjugated Descendent algorithm and returns the mean square error of neural network output compared to expected output.

Parameters:

nInputs - number of inputs in first layer

nOutputs - number of output in output layer

examples - training examples

outputs - expected outputs

topology - net topology (cardinality of the hidder layers)

weights - net weights

r - random generator

Returns:

the mean square error of neural network output compared to expected output.

nnoutput

public double[] nnoutput(double[] x)

Calculated the output of present perceptron with input x and returns it in original scale.

Parameters:

x - the inputs for feeding the perceptron.

Returns:

a vector with the output of perceptron.

printNet

public java.lang.String printNet()

Returns the weights as a String for printing or storing purposes

Returns:

the weights in a String