keel.Algorithms.Neural_Networks.IRPropPlus_Clas

Class MSEOptimizableSigmNeuralNetClassifier

java.lang.Object

keel.Algorithms.Neural_Networks.NNEP_Common.neuralnet.AbstractNeuralNet

keel.Algorithms.Neural_Networks.NNEP_Clas.neuralnet.NeuralNetClassifier

keel.Algorithms.Neural_Networks.IRPropPlus_Clas.MSEOptimizableSigmNeuralNetClassifier

All Implemented Interfaces:

IOptimizableFunc, IClassifier, ISoftmaxClassifier, INeuralNet

public class MSEOptimizableSigmNeuralNetClassifier
extends NeuralNetClassifier
implements IOptimizableFunc

Sigmoid Neural Net with only a hidden layer and multiple outputs (classifier) Prepared for optimizing MSE.

Since:

JDK1.5

Version:

0.1

Author:

Written by Alfonso Carlos Martinez Estudillo (University of Cordoba) 5/11/2007, Written by Pedro Antonio Gutiérrez Peña (University of Crodoba) 5/11/2007

Field Summary

Fields

Modifier and Type	Field and Description
protected double	lastError Wrapped algorithm

Fields inherited from class keel.Algorithms.Neural_Networks.NNEP_Common.neuralnet.AbstractNeuralNet

hiddenLayers, inputLayer, outputLayer, XML

Constructor Summary

Constructors

Constructor and Description
MSEOptimizableSigmNeuralNetClassifier()

Method Summary

Modifier and Type	Method and Description
double[]	getCoefficients() Returns the initial value of a[], that is, the coefficients of the model B01 B02 ...
double	getLastError() Last error of the model
double[]	gradient(double[][] x, double[][] y) Returns the gradient vector of the derivative of MSE error function with respect to each coefficient of the model, using an input observation matrix (x[]) and an expected output matrix (y[])
void	setCoefficients(double[] a) Establish the final value of a[], that is, the coefficients of model B01 B02 ...

Methods inherited from class keel.Algorithms.Neural_Networks.NNEP_Clas.neuralnet.NeuralNetClassifier

classify, classify, rawOutputs, rawOutputs, softmaxProbabilities, softmaxProbabilities

Methods inherited from class keel.Algorithms.Neural_Networks.NNEP_Common.neuralnet.AbstractNeuralNet

addHlayer, copy, equals, getHlayer, getInputLayer, getNofhlayers, getNofhneurons, getNoflinks, getOutputLayer, hashCode, keepRelevantLinks, linksEmpty, linksFull, neuronsEmpty, neuronsFull, setInputLayer, setOutputLayer, toString

Methods inherited from class java.lang.Object

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

Field Detail

lastError

protected double lastError

Wrapped algorithm

Constructor Detail

MSEOptimizableSigmNeuralNetClassifier

public MSEOptimizableSigmNeuralNetClassifier()

Method Detail

getCoefficients

public double[] getCoefficients()

Returns the initial value of a[], that is, the coefficients of the model B01 B02 ... B0(J-1) [W11 W12 ... B11 B12 B1(J-1)]* ...

Specified by:

getCoefficients in interface IOptimizableFunc

Returns:

double array of initial coefficients values

setCoefficients

public void setCoefficients(double[] a)

Establish the final value of a[], that is, the coefficients of model B01 B02 ... B0(J-1) [W11 W12 ... B11 B12 B1(J-1)]* ...

Specified by:

setCoefficients in interface IOptimizableFunc

Parameters:

a - array of final coefficients values

gradient

public double[] gradient(double[][] x,
                         double[][] y)

Returns the gradient vector of the derivative of MSE error function with respect to each coefficient of the model, using an input observation matrix (x[]) and an expected output matrix (y[])

Specified by:

gradient in interface IOptimizableFunc

Parameters:

x - Array with all inputs of all observations

y - Array with all expected outputs of all observations

Returns:

double[] Gradient vector gradient of dE/da for all coefficients

getLastError

public double getLastError()

Last error of the model

Specified by:

getLastError in interface IOptimizableFunc

Returns:

double Error of the function of the model with respect to data y[]