org.encog.neural.networks.training.propagation.back
Class BackpropagationMethod

java.lang.Object
  org.encog.neural.networks.training.propagation.back.BackpropagationMethod

All Implemented Interfaces:

PropagationMethod

public class BackpropagationMethod
extends java.lang.Object
implements PropagationMethod

This class implements the specifics of how the backpropagation algorithm is used. Specifically, the partial derivatives are simply applied to the weight matrix.

Author:

jheaton

Constructor Summary

BackpropagationMethod()

Method Summary

void	calculateError(NeuralOutputHolder output, PropagationLevel fromLevel, PropagationLevel toLevel) Calculate the error between these two levels.
void	init(Propagation propagation) Setup this propagation method using the specified propagation class.
void	learn() Modify the weight matrix and thresholds based on the last call to calcError.

Methods inherited from class java.lang.Object

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

Constructor Detail

BackpropagationMethod

public BackpropagationMethod()

Method Detail

calculateError

public void calculateError(NeuralOutputHolder output,
                           PropagationLevel fromLevel,
                           PropagationLevel toLevel)

Calculate the error between these two levels.

Specified by:

calculateError in interface PropagationMethod

Parameters:

output - The output to the "to level".

fromLevel - The from level.

toLevel - The target level.

init

public void init(Propagation propagation)

Setup this propagation method using the specified propagation class.

Specified by:

init in interface PropagationMethod

Parameters:

propagation - The propagation class creating this method.

learn

public void learn()

Modify the weight matrix and thresholds based on the last call to calcError.

Specified by:

learn in interface PropagationMethod