Package cc.mallet.fst

Class CRFTrainerByValueGradients

java.lang.Object

cc.mallet.fst.TransducerTrainer

cc.mallet.fst.CRFTrainerByValueGradients

All Implemented Interfaces:

TransducerTrainer.ByOptimization

public class CRFTrainerByValueGradients
extends TransducerTrainer
implements TransducerTrainer.ByOptimization

A CRF trainer that can combine multiple objective functions, each represented by a Optmizable.ByValueGradient.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
class	CRFTrainerByValueGradients.OptimizableCRF	An optimizable CRF that contains a collection of objective functions.

Nested classes/interfaces inherited from class cc.mallet.fst.TransducerTrainer

TransducerTrainer.ByIncrements, TransducerTrainer.ByInstanceIncrements, TransducerTrainer.ByOptimization

Field Summary

Fields

Modifier and Type	Field	Description
static int	DEFAULT_MAX_RESETS

Constructor Summary

Constructors

Constructor	Description
CRFTrainerByValueGradients(CRF crf, Optimizable.ByGradientValue[] optimizableByValueGradientObjects)

Method Summary

Modifier and Type	Method	Description
CRF	getCRF()
int	getIteration()
Optimizable.ByGradientValue[]	getOptimizableByGradientValueObjects()
CRFTrainerByValueGradients.OptimizableCRF	getOptimizableCRF(InstanceList trainingSet)	Returns an optimizable CRF that contains a collection of objective functions.
Optimizer	getOptimizer()
Optimizer	getOptimizer(InstanceList trainingSet)	Returns a L-BFGS optimizer, creating if one doesn't exist.
Transducer	getTransducer()
boolean	isConverged()	Returns true if training converged, false otherwise.
boolean	isFinishedTraining()	Returns true if training converged, false otherwise.
void	setMaxResets(int maxResets)	Sets the max.
boolean	train(InstanceList trainingSet, int numIterations)	Trains a CRF until convergence or specified number of iterations, whichever is earlier.
boolean	train(InstanceList training, int numIterationsPerProportion, double[] trainingProportions)	Train a CRF on various-sized subsets of the data.
boolean	trainIncremental(InstanceList training)	Trains a CRF until convergence.

Methods inherited from class cc.mallet.fst.TransducerTrainer

addEvaluator, addEvaluators, removeEvaluator, runEvaluators, train

Methods inherited from class java.lang.Object

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

Field Detail

DEFAULT_MAX_RESETS

public static final int DEFAULT_MAX_RESETS

See Also:

Constant Field Values

Constructor Detail

CRFTrainerByValueGradients

public CRFTrainerByValueGradients(CRF crf,
                                  Optimizable.ByGradientValue[] optimizableByValueGradientObjects)

Method Detail

getTransducer

public Transducer getTransducer()

Specified by:

getTransducer in class TransducerTrainer

getCRF

public CRF getCRF()

getOptimizer

public Optimizer getOptimizer()

Specified by:

getOptimizer in interface TransducerTrainer.ByOptimization

isConverged

public boolean isConverged()

Returns true if training converged, false otherwise.

isFinishedTraining

public boolean isFinishedTraining()

Returns true if training converged, false otherwise.

Specified by:

isFinishedTraining in class TransducerTrainer

getIteration

public int getIteration()

Specified by:

getIteration in class TransducerTrainer

getOptimizableByGradientValueObjects

public Optimizable.ByGradientValue[] getOptimizableByGradientValueObjects()

getOptimizableCRF

public CRFTrainerByValueGradients.OptimizableCRF getOptimizableCRF(InstanceList trainingSet)

Returns an optimizable CRF that contains a collection of objective functions.

If one doesn't exist then creates one and sets the optimizer to null.

getOptimizer

public Optimizer getOptimizer(InstanceList trainingSet)

Returns a L-BFGS optimizer, creating if one doesn't exist.

Also creates an optimizable CRF if required.

trainIncremental

public boolean trainIncremental(InstanceList training)

Trains a CRF until convergence.

train

public boolean train(InstanceList trainingSet,
                     int numIterations)

Trains a CRF until convergence or specified number of iterations, whichever is earlier.

Also creates an optimizable CRF and an optmizer if required.

Specified by:

train in class TransducerTrainer

train

public boolean train(InstanceList training,
                     int numIterationsPerProportion,
                     double[] trainingProportions)

Train a CRF on various-sized subsets of the data. This method is typically used to accelerate training by quickly getting to reasonable parameters on only a subset of the parameters first, then on progressively more data.

Parameters:

training - The training Instances.

numIterationsPerProportion - Maximum number of Maximizer iterations per training proportion.

trainingProportions - If non-null, train on increasingly larger portions of the data, e.g. new double[] {0.2, 0.5, 1.0}. This can sometimes speedup convergence. Be sure to end in 1.0 if you want to train on all the data in the end.

Returns:

True if training has converged.

setMaxResets

public void setMaxResets(int maxResets)

Sets the max. number of times the optimizer can be reset before throwing an exception.

Default value: DEFAULT_MAX_RESETS.