yawn.nn.mlp
Class PerceptronNode

java.lang.Object
  yawn.nn.NeuralNode
      yawn.nn.mlp.PerceptronNode

Direct Known Subclasses:

LinearNode, SigmoidNode

public abstract class PerceptronNode
extends NeuralNode

This is part of the yawn project. $Id: PerceptronNode.java,v 1.10 2005/04/07 17:28:07 supermarti Exp $

Version:

$Revision: 1.10 $

Author:

Luis Martí (luis dot marti at uc3m dot es)

Field Summary

protected double	maxWeightValue
protected double	minWeightValue
protected Pattern	momentum
protected Pattern	weights

Fields inherited from class yawn.nn.NeuralNode

input, inputSize

Constructor Summary

PerceptronNode()

PerceptronNode(int inputSize)

Method Summary

protected double	activationFunction(Pattern input)
void	adapt(double delta, double learningRate, double momentumRate) Adapt the weights using the delta rule.
double	calculateHiddenLayerDelta(Pattern nextLayerDeltas, Pattern nextLayerWeights) Calculate the deltas using the subsequent layer deltas and the weights emmanent from this node.
double	calculateOutputLayerDelta(double expected)
protected abstract double	df(double x)
protected abstract double	f(double x)
double	getMaxWeightValue()
double	getMinWeightValue()
Pattern	getWeights()
void	setInputSize(int inputSize)
void	setMaxWeightValue(double maxWeightValue)
void	setMinWeightValue(double minWeightValue)
void	setWeights(Pattern weights)

Methods inherited from class yawn.nn.NeuralNode

getInputSize, output, reset, setInput, setInputElement

Methods inherited from class java.lang.Object

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

Field Detail

weights

protected Pattern weights

momentum

protected Pattern momentum

minWeightValue

protected double minWeightValue

maxWeightValue

protected double maxWeightValue

Constructor Detail

PerceptronNode

public PerceptronNode(int inputSize)

Parameters:

inputSize -

PerceptronNode

public PerceptronNode()

Method Detail

setInputSize

public void setInputSize(int inputSize)

Overrides:

setInputSize in class NeuralNode

activationFunction

protected double activationFunction(Pattern input)

Specified by:

activationFunction in class NeuralNode

getWeights

public Pattern getWeights()

Returns:

the weigths of this node

setWeights

public void setWeights(Pattern weights)

Parameters:

weights -

f

protected abstract double f(double x)

df

protected abstract double df(double x)

adapt

public void adapt(double delta,
                  double learningRate,
                  double momentumRate)

Adapt the weights using the delta rule.

Parameters:

delta -

learningRate -

calculateOutputLayerDelta

public double calculateOutputLayerDelta(double expected)

calculateHiddenLayerDelta

public double calculateHiddenLayerDelta(Pattern nextLayerDeltas,
                                        Pattern nextLayerWeights)

Calculate the deltas using the subsequent layer deltas and the weights emmanent from this node.

Parameters:

nextLayerDeltas - the deltas of the next layer

nextLayerWeights - the weights of the connections emaning from this node

Returns:

the value of this node's delta

getMaxWeightValue

public double getMaxWeightValue()

Returns:

Returns the maxWeightValue.

setMaxWeightValue

public void setMaxWeightValue(double maxWeightValue)

Parameters:

maxWeightValue - The maxWeightValue to set.

getMinWeightValue

public double getMinWeightValue()

Returns:

Returns the minWeightValue.

setMinWeightValue

public void setMinWeightValue(double minWeightValue)

Parameters:

minWeightValue - The minWeightValue to set.