de.jstacs.models.discrete.inhomogeneous
Class StructureLearner

java.lang.Object
  de.jstacs.models.discrete.inhomogeneous.StructureLearner

public class StructureLearner
extends Object

This class can be used to learn the structure of any discrete model.

Author:

Jens Keilwagen

Nested Class Summary

static class	StructureLearner.LearningType This enum defines the different types of learning that are possible with the StructureLearner.
static class	StructureLearner.ModelType This enum defines the different types of models that can be learned with the StructureLearner.

Constructor Summary

StructureLearner(AlphabetContainer con, int length)
Creates a StructureLearner with ess = 0.

StructureLearner(AlphabetContainer con, int length, double ess)
Creates a StructureLearner

Method Summary

AlphabetContainer	getAlphabetContainer() This method returns the AlphabetContainer of the StructureLearner.
double	getEss() This method returns the ESS of the StructureLearner.
int[][]	getStructure(Sample data, double[] weights, StructureLearner.ModelType model, byte order, StructureLearner.LearningType method) This method finds the optimal structure (in some sense).
static int[][]	getStructure(Tensor t, StructureLearner.ModelType model, byte order) This method can be used to determine the optimal structure.
SymmetricTensor	getTensor(Sample data, double[] weights, byte order, StructureLearner.LearningType method) This method can be used to compute a tensor that can be used to determine the optimal structure.
void	setESS(double ess) This method sets the ESS of the StructureLearner.

Methods inherited from class java.lang.Object

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

Constructor Detail

StructureLearner

public StructureLearner(AlphabetContainer con,
                        int length,
                        double ess)
                 throws IllegalArgumentException

Creates a StructureLearner

Parameters:

con - the alphabets this instance should use

length - the length

ess - the ess (has to be non-negative)

Throws:

IllegalArgumentException - if the AlphabetContainer is not discrete, the length is not matching with the AlphabetContainer or the ess is below 0

StructureLearner

public StructureLearner(AlphabetContainer con,
                        int length)
                 throws IllegalArgumentException

Creates a StructureLearner with ess = 0.

Parameters:

con -

length -

Throws:

IllegalArgumentException - if the AlphabetContainer is not discrete or the length is not matching with the AlphabetContainer

See Also:

StructureLearner(AlphabetContainer, int, double)

Method Detail

getAlphabetContainer

public AlphabetContainer getAlphabetContainer()

This method returns the AlphabetContainer of the StructureLearner.

Returns:

the AlphabetContainer of the StructureLearner

getEss

public double getEss()

This method returns the ESS of the StructureLearner.

Returns:

the ESS of the StructureLearner

setESS

public void setESS(double ess)
            throws IllegalArgumentException

This method sets the ESS of the StructureLearner.

Parameters:

ess - the SS of the StructureLearner

Throws:

IllegalArgumentException - if the ess is below 0

getStructure

public int[][] getStructure(Sample data,
                            double[] weights,
                            StructureLearner.ModelType model,
                            byte order,
                            StructureLearner.LearningType method)
                     throws Exception

This method finds the optimal structure (in some sense).

Parameters:

data - the Sample

weights - the weights

model - the kind of model

order - the markov order

method - the learning method

Returns:

the structure of the specified model

Throws:

Exception - if something went wrong

getStructure

public static int[][] getStructure(Tensor t,
                                   StructureLearner.ModelType model,
                                   byte order)
                            throws Exception

This method can be used to determine the optimal structure.

Parameters:

t - the tensor containing all relevant weights

model - the model type

order - the model order

Returns:

the structure

Throws:

Exception - if something in the algorithm went wrong

getTensor

public SymmetricTensor getTensor(Sample data,
                                 double[] weights,
                                 byte order,
                                 StructureLearner.LearningType method)
                          throws IllegalArgumentException,
                                 WrongAlphabetException

This method can be used to compute a tensor that can be used to determine the optimal structure.

Parameters:

data - the data

weights - the weights

order - the markov order

method - the learning type

Returns:

a tensor containing all values the are necessary to determine the optimal structure

Throws:

IllegalArgumentException

WrongAlphabetException - if the AlphabetContainer of the data is not correct