org.encog.neural.networks.training.lma

Class LevenbergMarquardtTraining

java.lang.Object

org.encog.ml.train.BasicTraining

org.encog.neural.networks.training.lma.LevenbergMarquardtTraining

All Implemented Interfaces:

MLTrain, MultiThreadable

public class LevenbergMarquardtTraining
extends BasicTraining
implements MultiThreadable

Trains a neural network using a Levenberg Marquardt algorithm (LMA). This training technique is based on the mathematical technique of the same name. The LMA interpolates between the Gauss-Newton algorithm (GNA) and the method of gradient descent (similar to what is used by backpropagation. The lambda parameter determines the degree to which GNA and Gradient Descent are used. A lower lambda results in heavier use of GNA, whereas a higher lambda results in a heavier use of gradient descent. Each iteration starts with a low lambda that builds if the improvement to the neural network is not desirable. At some point the lambda is high enough that the training method reverts totally to gradient descent. This allows the neural network to be trained effectively in cases where GNA provides the optimal training time, but has the ability to fall back to the more primitive gradient descent method LMA finds only a local minimum, not a global minimum. References: http://www.heatonresearch.com/wiki/LMA http://en.wikipedia.org/wiki/Levenberg%E2%80%93Marquardt_algorithm http://en.wikipedia.org/wiki/Finite_difference_method http://crsouza.blogspot.com/2009/11/neural-network-learning-by-levenberg_18.html http://mathworld.wolfram.com/FiniteDifference.html http://www-alg.ist.hokudai.ac.jp/~jan/alpha.pdf - http://www.inference.phy.cam.ac.uk/mackay/Bayes_FAQ.html

Field Summary

Fields

Modifier and Type	Field and Description
static double	LAMBDA_MAX The max amount for the LAMBDA.
static double	SCALE_LAMBDA The amount to scale the lambda by.

Constructor Summary

Constructors

Constructor and Description
LevenbergMarquardtTraining(BasicNetwork network, MLDataSet training) Construct the LMA object.
LevenbergMarquardtTraining(BasicNetwork network, MLDataSet training, ComputeHessian h) Construct the LMA object.

Method Summary

Modifier and Type	Method and Description
boolean	canContinue()
ComputeHessian	getHessian()
MLMethod	getMethod() Get the current best machine learning method from the training.
int	getThreadCount()
void	iteration() Perform one iteration.
TrainingContinuation	pause() Pause the training to continue later.
void	resume(TrainingContinuation state) Resume training.
void	setThreadCount(int numThreads) Set the number of threads to use.
void	updateWeights() Update the weights in the neural network.

Methods inherited from class org.encog.ml.train.BasicTraining

addStrategy, finishTraining, getError, getImplementationType, getIteration, getStrategies, getTraining, isTrainingDone, iteration, postIteration, preIteration, setError, setIteration, setTraining

Methods inherited from class java.lang.Object

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

Field Detail

SCALE_LAMBDA

public static final double SCALE_LAMBDA

The amount to scale the lambda by.

See Also:

Constant Field Values

LAMBDA_MAX

public static final double LAMBDA_MAX

The max amount for the LAMBDA.

See Also:

Constant Field Values

Constructor Detail

LevenbergMarquardtTraining

public LevenbergMarquardtTraining(BasicNetwork network,
                                  MLDataSet training)

Construct the LMA object.

Parameters:

network - The network to train. Must have a single output neuron.

training - The training data to use. Must be indexable.

LevenbergMarquardtTraining

public LevenbergMarquardtTraining(BasicNetwork network,
                                  MLDataSet training,
                                  ComputeHessian h)

Construct the LMA object.

Parameters:

network - The network to train. Must have a single output neuron.

training - The training data to use. Must be indexable.

h - Utility class to compute the Hessian.

Method Detail

canContinue

public boolean canContinue()

Specified by:

canContinue in interface MLTrain

Returns:

True if the training can be paused, and later continued.

getMethod

public MLMethod getMethod()

Description copied from interface: MLTrain

Get the current best machine learning method from the training.

Specified by:

getMethod in interface MLTrain

Returns:

The trained network.

iteration

public void iteration()

Perform one iteration.

Specified by:

iteration in interface MLTrain

pause

public TrainingContinuation pause()

Pause the training to continue later.

Specified by:

pause in interface MLTrain

Returns:

A training continuation object.

resume

public void resume(TrainingContinuation state)

Resume training.

Specified by:

resume in interface MLTrain

Parameters:

state - The training continuation object to use to continue.

updateWeights

public void updateWeights()

Update the weights in the neural network.

getHessian

public ComputeHessian getHessian()

Returns:

The Hessian calculation method used.

getThreadCount

public int getThreadCount()

Specified by:

getThreadCount in interface MultiThreadable

Returns:

The number of threads to use, 0 to automatically determine based on core count.

setThreadCount

public void setThreadCount(int numThreads)

Description copied from interface: MultiThreadable

Set the number of threads to use.

Specified by:

setThreadCount in interface MultiThreadable

Parameters:

numThreads - The number of threads to use, or zero to automatically determine based on core count.