de.jstacs.classifiers
Class ClassifierFactory

java.lang.Object
  de.jstacs.classifiers.ClassifierFactory

public class ClassifierFactory
extends Object

This class allows to easily create classifiers from TrainableStatisticalModels, DifferentiableStatisticalModels, and DifferentiableSequenceScores. Most parameters of the classifiers are set to default values. If you like to set some of these additional parameters to non-standard values, directly use the constructors of TrainSMBasedClassifier, GenDisMixClassifier, or MSPClassifier.

Author:

Jan Grau

Constructor Summary

ClassifierFactory()

Method Summary

static AbstractClassifier	createClassifier(DifferentiableSequenceScore... models) Creates a classifier that is based on at least two DifferentiableSequenceScores.
static AbstractClassifier	createClassifier(double[] beta, DifferentiableStatisticalModel... models) Creates a classifier that is based on at least two DifferentiableStatisticalModels.
static AbstractClassifier	createClassifier(LearningPrinciple principle, DifferentiableStatisticalModel... models) Creates a classifier that is based on at least two DifferentiableStatisticalModels.
static AbstractClassifier	createGenerativeClassifier(TrainableStatisticalModel... models) Creates a classifier that is based on at least two TrainableStatisticalModels.

Methods inherited from class java.lang.Object

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

Constructor Detail

ClassifierFactory

public ClassifierFactory()

Method Detail

createGenerativeClassifier

public static AbstractClassifier createGenerativeClassifier(TrainableStatisticalModel... models)
                                                     throws IllegalArgumentException,
                                                            CloneNotSupportedException,
                                                            ClassDimensionException

Creates a classifier that is based on at least two TrainableStatisticalModels. Such models can be created using the TrainableStatisticalModelFactory or by directly using constructors of sub-classes of TrainableStatisticalModel. After the classifier has been created, it can be trained using the AbstractClassifier.train(de.jstacs.data.DataSet...) method using one data set for each TrainableStatisticalModel provided.

Parameters:

models - the models for the individual classes

Returns:

the classifier

Throws:

IllegalArgumentException - if the TrainableStatisticalModels do not describe a common domain of sequences

CloneNotSupportedException - if at least one TrainableStatisticalModel could not be cloned

ClassDimensionException - if the number of classes is below 2

createClassifier

public static AbstractClassifier createClassifier(LearningPrinciple principle,
                                                  DifferentiableStatisticalModel... models)
                                           throws IllegalArgumentException,
                                                  Exception

Creates a classifier that is based on at least two DifferentiableStatisticalModels. Such models can be created using the DifferentiableStatisticalModelFactory or by directly using constructors of sub-classes of DifferentiableStatisticalModel. After the classifier has been created, it can be trained using the AbstractClassifier.train(de.jstacs.data.DataSet...) method using one data set for each DifferentiableStatisticalModel provided. The LearningPrinciple (generative ML or MAP, or discriminative MCL or MSP) must be provided as first parameter.

Parameters:

principle - the learning principle

models - the models for the individual classes

Returns:

the classifier

Throws:

IllegalArgumentException - if the lengths of the DifferentiableStatisticalModels (see SequenceScore.getLength()) are incompatible

Exception - if something else went wrong

createClassifier

public static AbstractClassifier createClassifier(double[] beta,
                                                  DifferentiableStatisticalModel... models)
                                           throws IllegalArgumentException,
                                                  Exception

Creates a classifier that is based on at least two DifferentiableStatisticalModels. Such models can be created using the DifferentiableStatisticalModelFactory or by directly using constructors of sub-classes of DifferentiableStatisticalModel. After the classifier has been created, it can be trained using the AbstractClassifier.train(de.jstacs.data.DataSet...) method using one data set for each DifferentiableStatisticalModel provided. The beta array determines the employed learning principle and must be provided as first parameter.

Parameters:

beta - an array specifying the learning principle by weighting factors for conditional likelihood, likelihood and prior

models - the models for the individual classes

Returns:

the classifier

Throws:

IllegalArgumentException - if the lengths of the DifferentiableStatisticalModels (see SequenceScore.getLength()) are incompatible

Exception - if something else went wrong

See Also:

LearningPrinciple, LogGenDisMixFunction

createClassifier

public static AbstractClassifier createClassifier(DifferentiableSequenceScore... models)
                                           throws IllegalArgumentException,
                                                  Exception

Creates a classifier that is based on at least two DifferentiableSequenceScores. Such scores can be created using the DifferentiableStatisticalModelFactory (since DifferentiableStatisticalModels are also DifferentiableSequenceScores) or by directly using constructors of sub-classes of DifferentiableSequenceScore. After the classifier has been created, it can be trained using the discriminative MCL principle using the AbstractClassifier.train(de.jstacs.data.DataSet...) method using one data set for each DifferentiableSequenceScore provided.

Parameters:

models - the models for the individual classes

Returns:

the classifier

Throws:

IllegalArgumentException - if the lengths of the DifferentiableSequenceScores (see SequenceScore.getLength()) are incompatible

Exception - if something else went wrong