org.simbrain.network.trainers

Class LMSOffline

java.lang.Object

org.simbrain.network.trainers.Trainer

org.simbrain.network.trainers.LMSOffline

public class LMSOffline
extends Trainer

Offline/Batch Learning with least mean squares.

Author:

ztosi, jyoshimi

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	LMSOffline.SolutionType Solution methods for offline LMS.

Nested classes/interfaces inherited from class org.simbrain.network.trainers.Trainer

Trainer.DataNotInitializedException

Field Summary

Fields inherited from class org.simbrain.network.trainers.Trainer

network

Constructor Summary

Constructors

Constructor and Description
LMSOffline(Trainable network) Construct the LMSOOffline object, with a trainable network the Synapse group where the new synapses will be placed.

Method Summary

Modifier and Type	Method and Description
void	apply() Apply the algorithm.
double	getAlpha()
Randomizer	getNoiseGen()
ComboBoxWrapper	getSolutionType() Returns the current solution type inside a comboboxwrapper.
boolean	isNoiseAdded()
boolean	isRidgeRegression()
void	moorePenroseSolution(Trainable network) Moore penrose.
void	setAlpha(double alpha)
void	setNoiseAdded(boolean noiseAdded)
void	setRidgeRegression(boolean ridgeRegression)
void	setSolutionType(ComboBoxWrapper solutionType) Set the current parse style.
void	setSolutionType(LMSOffline.SolutionType solutionType) Set solution type.
void	weinerHopfSolution(Trainable network) Implements the Wiener-Hopf solution to LMS linear regression.

Methods inherited from class org.simbrain.network.trainers.Trainer

addListener, fireProgressUpdate, fireTrainingBegin, fireTrainingEnd, getListeners, getTrainable, getTrainableNetwork, removeListener, revalidateSynapseGroups

Methods inherited from class java.lang.Object

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

Constructor Detail

LMSOffline

public LMSOffline(Trainable network)

Construct the LMSOOffline object, with a trainable network the Synapse group where the new synapses will be placed.

Parameters:

network - the network to train

Method Detail

apply

public void apply()
           throws Trainer.DataNotInitializedException

Description copied from class: Trainer

Apply the algorithm.

Specified by:

apply in class Trainer

Throws:

Trainer.DataNotInitializedException - when input or target data have not been set.

weinerHopfSolution

public void weinerHopfSolution(Trainable network)

Implements the Wiener-Hopf solution to LMS linear regression. TODO: Fix progress updates to reflect actual training times & %s

Parameters:

network - the trainable network being trained

moorePenroseSolution

public void moorePenroseSolution(Trainable network)

Moore penrose.

Parameters:

network - the trainable network being trained

setSolutionType

public void setSolutionType(LMSOffline.SolutionType solutionType)

Set solution type.

Parameters:

solutionType - the solutionType to set

getSolutionType

public ComboBoxWrapper getSolutionType()

Returns the current solution type inside a comboboxwrapper. Used by preference dialog.

Returns:

the the comboBox

setSolutionType

public void setSolutionType(ComboBoxWrapper solutionType)

Set the current parse style. Used by preference dialog.

Parameters:

solutionType - the current solution.

isRidgeRegression

public boolean isRidgeRegression()

setRidgeRegression

public void setRidgeRegression(boolean ridgeRegression)

getAlpha

public double getAlpha()

setAlpha

public void setAlpha(double alpha)

isNoiseAdded

public boolean isNoiseAdded()

setNoiseAdded

public void setNoiseAdded(boolean noiseAdded)

getNoiseGen

public Randomizer getNoiseGen()