org.encog.neural.networks.training.svm
Class SVMTrain

java.lang.Object
  org.encog.neural.networks.training.BasicTraining
      org.encog.neural.networks.training.svm.SVMTrain

All Implemented Interfaces:

Train

public class SVMTrain
extends BasicTraining

Provides training for Support Vector Machine networks.

Field Summary

static double	DEFAULT_CONST_BEGIN The default starting number for C.
static double	DEFAULT_CONST_END The default ending number for C.
static double	DEFAULT_CONST_STEP The default step for C.
static double	DEFAULT_GAMMA_BEGIN The default gamma begin.
static double	DEFAULT_GAMMA_END The default gamma end.
static double	DEFAULT_GAMMA_STEP The default gamma step.

Constructor Summary

SVMTrain(BasicNetwork network, NeuralDataSet training)
Construct a trainer for an SVM network.

Method Summary

double	crossValidate(int index, double gamma, double c) Cross validate and check the specified index/gamma.
void	finishTraining() Called to finish training.
double	getConstBegin()
double	getConstEnd()
double	getConstStep()
int	getFold()
double	getGammaBegin()
double	getGammaEnd()
double	getGammaStep()
BasicNetwork	getNetwork() Get the current best network from the training.
svm_problem[]	getProblem()
boolean	isTrainingDone()
void	iteration() Perform one training iteration.
void	setConstBegin(double constBegin)
void	setConstEnd(double constEnd)
void	setConstStep(double constStep)
void	setFold(int fold)
void	setGammaBegin(double gammaBegin)
void	setGammaEnd(double gammaEnd)
void	setGammaStep(double gammaStep)
void	train() Quickly train all outputs with a C of 1.0 and a gamma equal to 1/(num inputs).
void	train(double gamma, double c) Quickly train the network with a fixed gamma and C.
void	train(int index, double gamma, double c) Quickly train one output with the specified gamma and C.

Methods inherited from class org.encog.neural.networks.training.BasicTraining

addStrategy, getCloud, getError, getIteration, getStrategies, getTraining, iteration, postIteration, preIteration, setCloud, setError, setIteration, setTraining

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_CONST_BEGIN

public static final double DEFAULT_CONST_BEGIN

The default starting number for C.

See Also:

Constant Field Values

DEFAULT_CONST_END

public static final double DEFAULT_CONST_END

The default ending number for C.

See Also:

Constant Field Values

DEFAULT_CONST_STEP

public static final double DEFAULT_CONST_STEP

The default step for C.

See Also:

Constant Field Values

DEFAULT_GAMMA_BEGIN

public static final double DEFAULT_GAMMA_BEGIN

The default gamma begin.

See Also:

Constant Field Values

DEFAULT_GAMMA_END

public static final double DEFAULT_GAMMA_END

The default gamma end.

See Also:

Constant Field Values

DEFAULT_GAMMA_STEP

public static final double DEFAULT_GAMMA_STEP

The default gamma step.

See Also:

Constant Field Values

Constructor Detail

SVMTrain

public SVMTrain(BasicNetwork network,
                NeuralDataSet training)

Construct a trainer for an SVM network.

Parameters:

network - The network to train.

training - The training data for this network.

Method Detail

train

public void train()

Quickly train all outputs with a C of 1.0 and a gamma equal to 1/(num inputs).

train

public void train(int index,
                  double gamma,
                  double c)

Quickly train one output with the specified gamma and C.

Parameters:

index - The output to train.

gamma - The gamma to train with.

c - The C to train with.

crossValidate

public double crossValidate(int index,
                            double gamma,
                            double c)

Cross validate and check the specified index/gamma.

Parameters:

index - The output index to cross validate.

gamma - The gamma to check.

c - The C to check.

Returns:

The calculated error.

iteration

public void iteration()

Perform one training iteration.

getProblem

public svm_problem[] getProblem()

Returns:

The problem being trained.

getFold

public int getFold()

Returns:

the fold

setFold

public void setFold(int fold)

Parameters:

fold - the fold to set

getConstBegin

public double getConstBegin()

Returns:

the constBegin

setConstBegin

public void setConstBegin(double constBegin)

Parameters:

constBegin - the constBegin to set

getConstStep

public double getConstStep()

Returns:

the constStep

setConstStep

public void setConstStep(double constStep)

Parameters:

constStep - the constStep to set

getConstEnd

public double getConstEnd()

Returns:

the constEnd

setConstEnd

public void setConstEnd(double constEnd)

Parameters:

constEnd - the constEnd to set

getGammaBegin

public double getGammaBegin()

Returns:

the gammaBegin

setGammaBegin

public void setGammaBegin(double gammaBegin)

Parameters:

gammaBegin - the gammaBegin to set

getGammaEnd

public double getGammaEnd()

Returns:

the gammaEnd

setGammaEnd

public void setGammaEnd(double gammaEnd)

Parameters:

gammaEnd - the gammaEnd to set

getGammaStep

public double getGammaStep()

Returns:

the gammaStep

setGammaStep

public void setGammaStep(double gammaStep)

Parameters:

gammaStep - the gammaStep to set

finishTraining

public void finishTraining()

Called to finish training.

Specified by:

finishTraining in interface Train

Overrides:

finishTraining in class BasicTraining

getNetwork

public BasicNetwork getNetwork()

Description copied from interface: Train

Get the current best network from the training.

Returns:

The trained network.

isTrainingDone

public boolean isTrainingDone()

Specified by:

isTrainingDone in interface Train

Overrides:

isTrainingDone in class BasicTraining

Returns:

True if the training is done.

train

public void train(double gamma,
                  double c)

Quickly train the network with a fixed gamma and C.

Parameters:

gamma - The gamma to use.

c - The C to use.