org.neuroph.core.learning

Class IterativeLearning

java.lang.Object

org.neuroph.core.learning.LearningRule

org.neuroph.core.learning.IterativeLearning

All Implemented Interfaces:

Serializable

Direct Known Subclasses:

SupervisedLearning, UnsupervisedLearning

public abstract class IterativeLearning
extends LearningRule
implements Serializable

Base class for all iterative learning algorithms. It provides the iterative learning procedure for all of its subclasses.

Author:

Zoran Sevarac

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected int	currentIteration Current iteration counter
protected boolean	iterationsLimited Flag for indicating if the training iteration number is limited
protected double	learningRate Learning rate parametar
protected int	maxIterations Max training iterations (when to stopLearning training) TODO: this field should be private, to force use of setMaxIterations from derived classes, so iterationsLimited flag is also set at the sam etime.Wil that break backward compatibility with serialized networks?

Fields inherited from class org.neuroph.core.learning.LearningRule

listeners, neuralNetwork

Constructor Summary

Constructors

Constructor and Description
IterativeLearning() Creates new instance of IterativeLearning learning algorithm

Method Summary

Methods

Modifier and Type	Method and Description
protected void	afterEpoch()
protected void	beforeEpoch()
abstract void	doLearningEpoch(DataSet trainingSet) Override this method to implement specific learning epoch - one learning iteration, one pass through whole training set
void	doOneLearningIteration(DataSet trainingSet) Runs one learning iteration for the specified training set and notfies observers.
Integer	getCurrentIteration() Returns current iteration of this learning algorithm
double	getLearningRate() Returns learning rate for this algorithm
boolean	isPausedLearning() Returns true if learning thread is paused, false otherwise
void	learn(DataSet trainingSet) Override this method to implement specific learning procedures
void	learn(DataSet trainingSet, int maxIterations) Trains network for the specified training set and number of iterations
protected void	onStart() This method is executed when learning starts, before the first epoch.
void	pause() Pause the learning
void	resume() Resumes the paused learning
void	setLearningRate(double learningRate) Sets learning rate for this algorithm
void	setMaxIterations(int maxIterations) Sets iteration limit for this learning algorithm

Methods inherited from class org.neuroph.core.learning.LearningRule

addListener, fireLearningEvent, getNeuralNetwork, getTrainingSet, isStopped, removeListener, setNeuralNetwork, setTrainingSet, stopLearning

Methods inherited from class java.lang.Object

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

Field Detail

learningRate

protected double learningRate

Learning rate parametar

currentIteration

protected int currentIteration

Current iteration counter

maxIterations

protected int maxIterations

Max training iterations (when to stopLearning training) TODO: this field should be private, to force use of setMaxIterations from derived classes, so iterationsLimited flag is also set at the sam etime.Wil that break backward compatibility with serialized networks?

iterationsLimited

protected boolean iterationsLimited

Flag for indicating if the training iteration number is limited

Constructor Detail

IterativeLearning

public IterativeLearning()

Creates new instance of IterativeLearning learning algorithm

Method Detail

getLearningRate

public double getLearningRate()

Returns learning rate for this algorithm

Returns:

learning rate for this algorithm

setLearningRate

public void setLearningRate(double learningRate)

Sets learning rate for this algorithm

Parameters:

learningRate - learning rate for this algorithm

setMaxIterations

public void setMaxIterations(int maxIterations)

Sets iteration limit for this learning algorithm

Parameters:

maxIterations - iteration limit for this learning algorithm

getCurrentIteration

public Integer getCurrentIteration()

Returns current iteration of this learning algorithm

Returns:

current iteration of this learning algorithm

isPausedLearning

public boolean isPausedLearning()

Returns true if learning thread is paused, false otherwise

Returns:

true if learning thread is paused, false otherwise

pause

public void pause()

Pause the learning

resume

public void resume()

Resumes the paused learning

onStart

protected void onStart()

This method is executed when learning starts, before the first epoch. Used for initialisation.

Overrides:

onStart in class LearningRule

beforeEpoch

protected void beforeEpoch()

afterEpoch

protected void afterEpoch()

learn

public final void learn(DataSet trainingSet)

Description copied from class: LearningRule

Override this method to implement specific learning procedures

Specified by:

learn in class LearningRule

Parameters:

trainingSet - training set

learn

public void learn(DataSet trainingSet,
         int maxIterations)

Trains network for the specified training set and number of iterations

Parameters:

trainingSet - training set to learn

maxIterations - maximum numberof iterations to learn

doOneLearningIteration

public void doOneLearningIteration(DataSet trainingSet)

Runs one learning iteration for the specified training set and notfies observers. This method does the the doLearningEpoch() and in addtion notifes observrs when iteration is done.

Parameters:

trainingSet - training set to learn

doLearningEpoch

public abstract void doLearningEpoch(DataSet trainingSet)

Override this method to implement specific learning epoch - one learning iteration, one pass through whole training set

Parameters:

trainingSet - training set