org.cognity.neural.train

Class GradientDescent

java.lang.Object

org.cognity.neural.train.GradientDescent

All Implemented Interfaces:

IterativeProcess, TrainingRule

Direct Known Subclasses:

Backpropagation

public abstract class GradientDescent
extends Object
implements TrainingRule, IterativeProcess

The base class for gradient descent algorithms.

Author:

Artur Markiewicz

Field Summary

Fields

Modifier and Type	Field and Description
static double	DEFAULT_LEARNING_RATE The default learning rate.
static double	DEFAULT_MOMENTUM The default momentum.

Constructor Summary

Constructors

Constructor and Description
GradientDescent() Creates a new gradient descent algorithm using default learning rate and momentum values.

Method Summary

Methods

Modifier and Type	Method and Description
double	getCurrentError() Returns the error value after the last iteration.
int	getCurrentIteration() Returns the current iteration.
double	getLastError() Returns the error value before the last iteration.
double	getLearningRate() Returns the learning rate value.
double	getMomentum() Returns the momentum value.
StopCondition	getStopCondition() Returns a stop condition for this process.
protected void	setCurrentError(double error) Sets the current error value.
void	setLearningRate(double rate) Sets the learning rate value.
void	setMomentum(double m) Sets the momentum value.
void	setStopCondition(StopCondition cond) Sets a stop condition for this process.
void	train(DataSet set) Performs the training of a neural network.
protected abstract void	trainExample(DataExample example) Learn a single data example.

Methods inherited from class java.lang.Object

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

Methods inherited from interface org.cognity.neural.train.TrainingRule

getNetwork

Field Detail

DEFAULT_LEARNING_RATE

public static final double DEFAULT_LEARNING_RATE

The default learning rate.

See Also:

Constant Field Values

DEFAULT_MOMENTUM

public static final double DEFAULT_MOMENTUM

The default momentum.

See Also:

Constant Field Values

Constructor Detail

GradientDescent

public GradientDescent()

Creates a new gradient descent algorithm using default learning rate and momentum values.

Method Detail

getLearningRate

public final double getLearningRate()

Returns the learning rate value.

Returns:

the learning rate value

setLearningRate

public final void setLearningRate(double rate)

Sets the learning rate value.

Parameters:

rate - the learning rate value. Must be a positive real number.

getMomentum

public final double getMomentum()

Returns the momentum value.

Returns:

the momentum value

setMomentum

public final void setMomentum(double m)

Sets the momentum value.

Parameters:

m - the momentum. Must be a non-negative real number.

getCurrentError

public final double getCurrentError()

Returns the error value after the last iteration.

Returns:

the error value after the last iteration

getLastError

public final double getLastError()

Returns the error value before the last iteration.

Returns:

the error value before the last iteration

train

public final void train(DataSet set)

Description copied from interface: TrainingRule

Performs the training of a neural network.

Specified by:

train in interface TrainingRule

Parameters:

set - training data set

getCurrentIteration

public final int getCurrentIteration()

Description copied from interface: IterativeProcess

Returns the current iteration.

Specified by:

getCurrentIteration in interface IterativeProcess

Returns:

the current iteration

setStopCondition

public final void setStopCondition(StopCondition cond)

Description copied from interface: IterativeProcess

Sets a stop condition for this process. Process should stop after the condition is met.

Specified by:

setStopCondition in interface IterativeProcess

Parameters:

cond - the stop condition. Cannot be null.

getStopCondition

public final StopCondition getStopCondition()

Description copied from interface: IterativeProcess

Returns a stop condition for this process.

Specified by:

getStopCondition in interface IterativeProcess

Returns:

a stop condition

trainExample

protected abstract void trainExample(DataExample example)

Learn a single data example.

Parameters:

example -

setCurrentError

protected final void setCurrentError(double error)

Sets the current error value.

Parameters:

error -