org.neuroph.contrib.samples

Class SunSpots

java.lang.Object

org.neuroph.contrib.samples.SunSpots

All Implemented Interfaces:

java.util.EventListener, LearningEventListener

public class SunSpots
extends java.lang.Object
implements LearningEventListener

This example shows how to use Neuroph to predict sunspots. It demonstrates two very important machine learning techniques. First, time-window. Sunspots are organized into input windows used to predict the next level of sunspot activity. This example uses a 30 year window. Basically, 30 years of sunspot activity is used to predict the 31st year. This 30 year window slides forward, one year at a time. Second is normalization. The sunspots are normalized into a range between 0.1 and 0.9. This is very close to the actual 0 to 1 range of the sigmoid function. We stay away from the extream edges of this range, thus using 0.1 and 0.9. This is an example from the book "Introduction to Neural Networks for Java" by Jeff Heaton. This example has been contributed to the Neuroph project by Jeff Heaton. http://www.heatonresearch.com/book/programming-neural-networks-java-2.html

Field Summary

Fields

Modifier and Type	Field and Description
static int	EVALUATE_END End of evaluation data.
static int	EVALUATE_START Beginning of evaluation data.
static double	MAX_ERROR This really should be lowered, I am setting it to a level here that will train in under a minute.
static int	STARTING_YEAR Starting year for sunspot data.
static double[]	SUNSPOTS
static int	TRAIN_END End of training data.
static int	TRAIN_START Start of training data.
static int	WINDOW_SIZE Size of our prediction window.

Constructor Summary

Constructors

Constructor and Description
SunSpots()

Method Summary

Modifier and Type	Method and Description
DataSet	generateTrainingData() Generate the training data for the training sunspot years.
void	handleLearningEvent(LearningEvent event) This method gets executed when LearningRule fires LearningEvent which some class is listening to.
static void	main(java.lang.String[] args)
void	normalizeSunspots(double lo, double hi) Normalize the sunspots.
void	predict(NeuralNetwork network) Predict sunspots.
void	run()

Methods inherited from class java.lang.Object

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

Field Detail

SUNSPOTS

public static final double[] SUNSPOTS

STARTING_YEAR

public static final int STARTING_YEAR

Starting year for sunspot data.

See Also:

Constant Field Values

WINDOW_SIZE

public static final int WINDOW_SIZE

Size of our prediction window.

See Also:

Constant Field Values

TRAIN_START

public static final int TRAIN_START

Start of training data.

See Also:

Constant Field Values

TRAIN_END

public static final int TRAIN_END

End of training data.

See Also:

Constant Field Values

EVALUATE_START

public static final int EVALUATE_START

Beginning of evaluation data.

See Also:

Constant Field Values

EVALUATE_END

public static final int EVALUATE_END

End of evaluation data.

MAX_ERROR

public static final double MAX_ERROR

This really should be lowered, I am setting it to a level here that will train in under a minute.

See Also:

Constant Field Values

Constructor Detail

SunSpots

public SunSpots()

Method Detail

normalizeSunspots

public void normalizeSunspots(double lo,
                              double hi)

Normalize the sunspots.

Parameters:

lo - Low range for normalization.

hi - High range for normalization.

generateTrainingData

public DataSet generateTrainingData()

Generate the training data for the training sunspot years.

Returns:

The training data.

predict

public void predict(NeuralNetwork network)

Predict sunspots.

Parameters:

network - Neural network to use.

run

public void run()

main

public static void main(java.lang.String[] args)

handleLearningEvent

public void handleLearningEvent(LearningEvent event)

Description copied from interface: LearningEventListener

This method gets executed when LearningRule fires LearningEvent which some class is listening to. For example, if you want to print current iteration, error etc.

Specified by:

handleLearningEvent in interface LearningEventListener

Parameters:

event - holds the information about event tha occured