org.joone.engine
Class Monitor

java.lang.Object
  org.joone.engine.Monitor

All Implemented Interfaces:

java.io.Serializable

public class Monitor
extends java.lang.Object
implements java.io.Serializable

The Monitor object is the controller of the behavior of the neural net. It controls the start/stop actions and permits to set the parameters of the net (learning rate, momentum, ecc.). Each component of the neural net (Layers and Synapses) are connected to a monitor object (the monitor can be different or the same for the all components).

See Also:

Serialized Form

Field Summary

protected boolean	backprop Ability to switch on and off the error backpropagation method, specifically for use with RTRL
int	learningMode

Constructor Summary

Monitor()
This is the default Constructor.

Method Summary

void	addLearner(int i, java.lang.String learner) Adds a new Learner to the Neural Network Usage: Monitor.addLearner(0, "org.joone.engine.BasicLearner"); Monitor.addLearner(1, "org.joone.engine.BatchLearner"); Monitor.addLearner(2, "org.joone.engine.RpropLearner");
void	addNeuralNetListener(NeuralNetListener l) adds a neural net event listener the Monitor
void	addNeuralNetListener(NeuralNetListener l, boolean removable) adds a neural net event listener to the Monitor
java.util.TreeSet	check()
void	fireCicleTerminated() Invoked when an epoch finishes
void	fireErrorChanged() Invoked when the GlobalError changes
void	fireErrorPattern()
void	fireNetStarted() Invoked when the net starts
void	fireNetStopped() Invoked when all the epochs finish
void	fireNetStoppedError(java.lang.String errMsg) Invoked when an error occurs
int	getBatchSize() Getter for the property BatchSize
int	getCurrentCicle() Returns the current epoch
double	getGlobalError() Returns the actual (R)MSE of the NN
java.lang.String[]	getKeys() Return all the keys of the parameters contained in this Monitor.
Learner	getLearner() Gets a learner for a synapse or layer.
Learner	getLearner(int index) Getter for the Learner declared at position 'index'
protected java.util.List	getLearners() Getter for property learners.
int	getLearningMode() Getter for property learningMode.
double	getLearningRate() Returns the learning rate
double	getMomentum() Returns the momentum
int	getNumOfPatterns()
java.lang.Object	getParam(java.lang.String key) Gets a custom parameter from the Monitor.
Monitor	getParent() Getter for property parent.
int	getPreLearning() Returns the initial ignored input patterns (during the training phase)
int	getRun() Needed for XML saving
int	getStep() Returns the actual elaborated pattern
int	getTotCicles() Returns the total number of epochs
int	getTrainingPatterns() Returns the number of the input training patterns
int	getValidationPatterns() Returns the number of the input validation patterns
void	Go() Deprecated.
boolean	isBackprop() See if the backprop is enabled
boolean	isExporting() Getter for property exporting.
boolean	isLearning() Returns TRUE if the net is into a learning phase
boolean	isLearningCicle(int num) Returns the phase of the net (learning or not) for the current pattern
boolean	isSingleThreadMode()
boolean	isSupervised() Getter for property supervised.
boolean	isTrainingDataForValidation() Getter for the property trainingDataForValidation, i.e.
boolean	isUseRMSE() Checks if we should use RMSE (true) or MSE false.
boolean	isValidation() Getter for property validation.
boolean	nextStep() Returns if the next pattern must be elaborated
protected java.lang.Object	readResolve()
void	removeAllListeners() Removes all the NeuralNetListeners.
void	removeNeuralNetListener(NeuralNetListener l) Removes a listener
void	resetCycle()
void	runAgain() Let continue the net.
void	setBackprop(boolean backprop) Turns backpropagation on or off
void	setBatchSize(int i) Setter for the property BatchSize
void	setCurrentCicle(int newCurrentCicle) Not used
void	setExporting(boolean exporting) Setter for property exporting.
void	setGlobalError(double newGlobalError) Sets the actual error of the NN
void	setLearnerFactory(LearnerFactory aFactory) Set learner factory.
protected void	setLearners(java.util.List learners) Setter for property learners.
void	setLearning(boolean newLearning) Sets the phase of the neural network: learning (true) or recall (false)
void	setLearningMode(int learningMode) Setter for property learningMode.
void	setLearningRate(double newLearningRate) Sets the learning rate
void	setMomentum(double newMomentum) Sets the momentum
void	setParam(java.lang.String key, java.lang.Object obj) Sets a custom parameter of the Monitor.
void	setParent(Monitor parent) Setter for property parent.
void	setPreLearning(int newPreLearning) Sets the initial ignored input patterns (during the training phase)
void	setSingleThreadMode(boolean singleThreadMode)
void	setSupervised(boolean supervised) Setter for property supervised.
void	setTotCicles(int newTotCicles) Sets the total number of epochs
void	setTrainingDataForValidation(boolean aMode) Setter for the property trainingDataForValidation.
void	setTrainingPatterns(int newPatterns) Sets the number of the input training patterns
void	setUseRMSE(boolean aMode) Uses the RMSE when set to true.
void	setValidation(boolean validation) Setter for property validation.
void	setValidationPatterns(int newPatterns) Sets the number of the input validation patterns
void	Stop() Stops the NN when running in multi-thread mode.

Methods inherited from class java.lang.Object

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

Field Detail

learningMode

public int learningMode

backprop

protected boolean backprop

Ability to switch on and off the error backpropagation method, specifically for use with RTRL

Constructor Detail

Monitor

public Monitor()

This is the default Constructor.

Method Detail

addNeuralNetListener

public void addNeuralNetListener(NeuralNetListener l)

adds a neural net event listener the Monitor

Parameters:

l - NeuralNetListener

addNeuralNetListener

public void addNeuralNetListener(NeuralNetListener l,
                                 boolean removable)

adds a neural net event listener to the Monitor

Parameters:

l - the new NeuralNetListener

removable - true if the added listener can be removed by the removeAllListeners method call

fireCicleTerminated

public void fireCicleTerminated()

Invoked when an epoch finishes

fireErrorPattern

public void fireErrorPattern()

fireNetStarted

public void fireNetStarted()

Invoked when the net starts

fireNetStopped

public void fireNetStopped()

Invoked when all the epochs finish

fireNetStoppedError

public void fireNetStoppedError(java.lang.String errMsg)

Invoked when an error occurs

Parameters:

errMsg - the thrown error message

fireErrorChanged

public void fireErrorChanged()

Invoked when the GlobalError changes

getCurrentCicle

public int getCurrentCicle()

Returns the current epoch

Returns:

int

getGlobalError

public double getGlobalError()

Returns the actual (R)MSE of the NN

Returns:

double

getLearningRate

public double getLearningRate()

Returns the learning rate

Returns:

double

getMomentum

public double getMomentum()

Returns the momentum

Returns:

double

getTrainingPatterns

public int getTrainingPatterns()

Returns the number of the input training patterns

Returns:

int

setTrainingPatterns

public void setTrainingPatterns(int newPatterns)

Sets the number of the input training patterns

Parameters:

newPatterns - int

getPreLearning

public int getPreLearning()

Returns the initial ignored input patterns (during the training phase)

Returns:

int

getStep

public int getStep()

Returns the actual elaborated pattern

Returns:

int

getTotCicles

public int getTotCicles()

Returns the total number of epochs

Returns:

int

Go

@Deprecated
public void Go()

Deprecated.

Runs the neural net in multi-thread mode. WARNING: AVOID to invoke this method. Use instead NeuralNet.go()

See Also:

org.joone.net.NeuralNet.go()

isLearning

public boolean isLearning()

Returns TRUE if the net is into a learning phase

Returns:

boolean

isLearningCicle

public boolean isLearningCicle(int num)

Returns the phase of the net (learning or not) for the current pattern

Parameters:

num - the pattern requested

Returns:

boolean

resetCycle

public void resetCycle()

nextStep

public boolean nextStep()

Returns if the next pattern must be elaborated

Returns:

boolean

readResolve

protected java.lang.Object readResolve()

removeNeuralNetListener

public void removeNeuralNetListener(NeuralNetListener l)

Removes a listener

Parameters:

l - the listener to be removed

runAgain

public void runAgain()

Let continue the net.

setCurrentCicle

public void setCurrentCicle(int newCurrentCicle)

Not used

Parameters:

newCurrentCicle - int

setGlobalError

public void setGlobalError(double newGlobalError)

Sets the actual error of the NN

Parameters:

newGlobalError - double

setLearning

public void setLearning(boolean newLearning)

Sets the phase of the neural network: learning (true) or recall (false)

Parameters:

newLearning - boolean

setLearningRate

public void setLearningRate(double newLearningRate)

Sets the learning rate

Parameters:

newLearningRate - double

setMomentum

public void setMomentum(double newMomentum)

Sets the momentum

Parameters:

newMomentum - double

setPreLearning

public void setPreLearning(int newPreLearning)

Sets the initial ignored input patterns (during the training phase)

Parameters:

newPreLearning - int

setTotCicles

public void setTotCicles(int newTotCicles)

Sets the total number of epochs

Parameters:

newTotCicles - int

Stop

public void Stop()

Stops the NN when running in multi-thread mode. WARNING: DO NOT INVOKE directly, use instead NeuralNet.stop()

See Also:

org.joone.net.NeuralNet.stop()

isExporting

public boolean isExporting()

Getter for property exporting.

Returns:

Value of property exporting.

setExporting

public void setExporting(boolean exporting)

Setter for property exporting.

Parameters:

exporting - New value of property exporting.

getRun

public int getRun()

Needed for XML saving

isValidation

public boolean isValidation()

Getter for property validation.

Returns:

Value of property validation.

setValidation

public void setValidation(boolean validation)

Setter for property validation.

Parameters:

validation - New value of property validation.

isTrainingDataForValidation

public boolean isTrainingDataForValidation()

Getter for the property trainingDataForValidation, i.e. should the training data be used for validation.

Returns:

true if the training data should be used, false otherwise.

setTrainingDataForValidation

public void setTrainingDataForValidation(boolean aMode)

Setter for the property trainingDataForValidation.

Parameters:

aMode - true if the training data should be used for validation, false otherwise.

removeAllListeners

public void removeAllListeners()

Removes all the NeuralNetListeners. Removes only the listeners marked as removable

See Also:

addNeuralNetListener

getParent

public Monitor getParent()

Getter for property parent.

Returns:

Value of property parent.

setParent

public void setParent(Monitor parent)

Setter for property parent.

Parameters:

parent - New value of property parent.

getValidationPatterns

public int getValidationPatterns()

Returns the number of the input validation patterns

Returns:

int

setValidationPatterns

public void setValidationPatterns(int newPatterns)

Sets the number of the input validation patterns

Parameters:

newPatterns - int

getNumOfPatterns

public int getNumOfPatterns()

check

public java.util.TreeSet check()

isSupervised

public boolean isSupervised()

Getter for property supervised.

Returns:

Value of property supervised.

setSupervised

public void setSupervised(boolean supervised)

Setter for property supervised. (default = false)

Parameters:

supervised - New value of property supervised.

getBatchSize

public int getBatchSize()

Getter for the property BatchSize

Returns:

the size (# of cycles) of the batch mode

setBatchSize

public void setBatchSize(int i)

Setter for the property BatchSize

Parameters:

the - size (# of cycles) of the batch mode

getLearningMode

public int getLearningMode()

Getter for property learningMode. The learningMode determines the kind of learning algorithm applied to the neural network.

Returns:

Value of property learningMode.

See Also:

getLearners()

setLearningMode

public void setLearningMode(int learningMode)

Setter for property learningMode.

Parameters:

learningMode - New value of property learningMode.

getLearner

public Learner getLearner(int index)

Getter for the Learner declared at position 'index'

Parameters:

the - index of the Learner to get

Returns:

the Learner at 'index' position

See Also:

getLearners()

getLearner

public Learner getLearner()

Gets a learner for a synapse or layer.

Returns:

a learner for a synapse or layer.

getLearners

protected java.util.List getLearners()

Getter for property learners.

Returns:

Value of property learners.

setLearners

protected void setLearners(java.util.List learners)

Setter for property learners. Used to set all the Learner objects used by this NN.

Parameters:

learners - New value of property learners.

addLearner

public void addLearner(int i,
                       java.lang.String learner)

Adds a new Learner to the Neural Network Usage: Monitor.addLearner(0, "org.joone.engine.BasicLearner"); Monitor.addLearner(1, "org.joone.engine.BatchLearner"); Monitor.addLearner(2, "org.joone.engine.RpropLearner");

Parameters:

i - the index of the new Learner

learner - a String containing the class name of the Learner to add

getParam

public java.lang.Object getParam(java.lang.String key)

Gets a custom parameter from the Monitor. The user is free to use the custom parameters as s/he wants. They are useful to store a whatever value that could be used by the nnet's components. It has been introduced to set the parameters of new added Learners.

Parameters:

key - The searched key

Returns:

The value of the parameter if found, otherwise null

setParam

public void setParam(java.lang.String key,
                     java.lang.Object obj)

Sets a custom parameter of the Monitor. The user is free to use the custom parameters as s/he wants. They are useful to store a whatever value that could be used by the nnet's components. It has been introduced to set the parameters of new added Learners.

Parameters:

key - The key of the param

obj - The value of the param

getKeys

public java.lang.String[] getKeys()

Return all the keys of the parameters contained in this Monitor.

Returns:

An array of Strings containing all the keys if found, otherwise null

setUseRMSE

public void setUseRMSE(boolean aMode)

Uses the RMSE when set to true. Uses MSE when set to false.

Parameters:

aMode - the mode to set true for RMSE, false for MSE.

isUseRMSE

public boolean isUseRMSE()

Checks if we should use RMSE (true) or MSE false.

Returns:

true if we should use RMSE, false if we should use MSE.

setLearnerFactory

public void setLearnerFactory(LearnerFactory aFactory)

Set learner factory.

Parameters:

aFactory - the learner factory to set.

isSingleThreadMode

public boolean isSingleThreadMode()

setSingleThreadMode

public void setSingleThreadMode(boolean singleThreadMode)

isBackprop

public boolean isBackprop()

See if the backprop is enabled

setBackprop

public void setBackprop(boolean backprop)

Turns backpropagation on or off