Package org.aksw.limes.core.ml.algorithm.dragon

Class Dragon

java.lang.Object

org.aksw.limes.core.ml.algorithm.ACoreMLAlgorithm

org.aksw.limes.core.ml.algorithm.dragon.Dragon

public class Dragon
extends ACoreMLAlgorithm

This class uses decision trees and an active learning approach to learn link specifications

Author:

Daniel Obraczka < soz11ffe@ studserv.uni-leipzig.de>

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
class	Dragon.SourceTargetValue	Helper class for easier handling of links or link candidates

Field Summary

Fields

Modifier and Type	Field	Description
static Set<String>	defaultMeasures
static String	FITNESS_NAME_GINI_INDEX
static String	FITNESS_NAME_GLOBAL_FMEASURE
static String	PARAMETER_ATOMIC_MEASURES
static String	PARAMETER_FITNESS_FUNCTION
static String	PARAMETER_MAPPING
static String	PARAMETER_MAX_LINK_SPEC_HEIGHT
static String	PARAMETER_MIN_PROPERTY_COVERAGE
static String	PARAMETER_PROPERTY_LEARNING_RATE
static String	PARAMETER_PROPERTY_MAPPING
static String	PARAMETER_PRUNING_CONFIDENCE
static String	PARAMETER_PRUNING_FUNCTION
static String	PRUNING_NAME_ERROR_ESTIMATE_PRUNING
static String	PRUNING_NAME_GLOBAL_FMEASURE
DecisionTree	root
static double	threshold

Fields inherited from class org.aksw.limes.core.ml.algorithm.ACoreMLAlgorithm

learningParameters, sourceCache, targetCache

Constructor Summary

Constructors

Constructor	Description
Dragon()	Dragon constructor.
Dragon(Configuration c)	Constructor uses superconstructor, initializes TreeParser object and sets configuration

Method Summary

Modifier and Type	Method	Description
protected MLResults	activeLearn()	generates an initial training set and calls activeLearn(AMapping)
protected MLResults	activeLearn(AMapping oracleMapping)	Creates a training set out of the oracleMapping and uses J48 to build a decision tree The decision tree gets parsed to a LinkSpecification
void	getAtomicMeasures()
Configuration	getConfiguration()
MLResults	getMlresult()
String	getName()	Name of the core ML algorithm.
protected AMapping	getNextExamples(int size)	Get a set of examples to be added to the mapping.
ACache	getSourceCache()
ACache	getTargetCache()
ACache	getTestSourceCache()
ACache	getTestTargetCache()
void	init(List<LearningParameter> lp, ACache sourceCache, ACache targetCache)	Initialize the core ML algorithm.
protected MLResults	learn(PseudoFMeasure pfm)	Learning method for unsupervised core ML algorithm implementations.
protected MLResults	learn(AMapping trainingData)	Learning method for supervised core ML algorithm implementations, where the confidence values for each pair in the trainingData determine its truth degree.
AMapping	predict(ACache source, ACache target, MLResults mlModel)	Predict/generate links from source to target based on mlModel.
void	setConfiguration(Configuration configuration)
void	setDefaultParameters()	Set default ACoreMLAlgorithm parameters values
void	setInitialMapping(AMapping initialMapping)
void	setSourceCache(ACache sourceCache)
void	setTargetCache(ACache targetCache)
protected boolean	supports(MLImplementationType mlType)	Check whether the mlType is supported.

Methods inherited from class org.aksw.limes.core.ml.algorithm.ACoreMLAlgorithm

getParameter, getParameters, setParameter

Methods inherited from class java.lang.Object

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

Field Detail

PARAMETER_PRUNING_CONFIDENCE

public static final String PARAMETER_PRUNING_CONFIDENCE

See Also:

Constant Field Values

PARAMETER_PROPERTY_MAPPING

public static final String PARAMETER_PROPERTY_MAPPING

See Also:

Constant Field Values

PARAMETER_MAPPING

public static final String PARAMETER_MAPPING

See Also:

Constant Field Values

PARAMETER_MAX_LINK_SPEC_HEIGHT

public static final String PARAMETER_MAX_LINK_SPEC_HEIGHT

See Also:

Constant Field Values

PARAMETER_MIN_PROPERTY_COVERAGE

public static final String PARAMETER_MIN_PROPERTY_COVERAGE

See Also:

Constant Field Values

PARAMETER_PROPERTY_LEARNING_RATE

public static final String PARAMETER_PROPERTY_LEARNING_RATE

See Also:

Constant Field Values

PARAMETER_FITNESS_FUNCTION

public static final String PARAMETER_FITNESS_FUNCTION

See Also:

Constant Field Values

PARAMETER_PRUNING_FUNCTION

public static final String PARAMETER_PRUNING_FUNCTION

See Also:

Constant Field Values

PARAMETER_ATOMIC_MEASURES

public static final String PARAMETER_ATOMIC_MEASURES

See Also:

Constant Field Values

FITNESS_NAME_GINI_INDEX

public static final String FITNESS_NAME_GINI_INDEX

See Also:

Constant Field Values

FITNESS_NAME_GLOBAL_FMEASURE

public static final String FITNESS_NAME_GLOBAL_FMEASURE

See Also:

Constant Field Values

PRUNING_NAME_ERROR_ESTIMATE_PRUNING

public static final String PRUNING_NAME_ERROR_ESTIMATE_PRUNING

See Also:

Constant Field Values

PRUNING_NAME_GLOBAL_FMEASURE

public static final String PRUNING_NAME_GLOBAL_FMEASURE

See Also:

Constant Field Values

root

public DecisionTree root

defaultMeasures

public static Set<String> defaultMeasures

threshold

public static final double threshold

See Also:

Constant Field Values

Constructor Detail

Dragon

public Dragon()

Dragon constructor.

Dragon

public Dragon(Configuration c)

Constructor uses superconstructor, initializes TreeParser object and sets configuration

Parameters:

c -

Method Detail

getName

public String getName()

Description copied from class: ACoreMLAlgorithm

Name of the core ML algorithm.

Specified by:

getName in class ACoreMLAlgorithm

Returns:

Name of the core ML algorithm.

activeLearn

protected MLResults activeLearn()
                         throws UnsupportedMLImplementationException

generates an initial training set and calls activeLearn(AMapping)

Specified by:

activeLearn in class ACoreMLAlgorithm

Returns:

wrap with results

Throws:

UnsupportedMLImplementationException - if ML implementation is not supported

getAtomicMeasures

public void getAtomicMeasures()

activeLearn

protected MLResults activeLearn(AMapping oracleMapping)
                         throws UnsupportedMLImplementationException

Creates a training set out of the oracleMapping and uses J48 to build a decision tree The decision tree gets parsed to a LinkSpecification

Specified by:

activeLearn in class ACoreMLAlgorithm

Parameters:

oracleMapping -

Returns:

res wrapper containing learned link specification

Throws:

UnsupportedMLImplementationException - if ML implementation is not supported

predict

public AMapping predict(ACache source,
                        ACache target,
                        MLResults mlModel)

Description copied from class: ACoreMLAlgorithm

Predict/generate links from source to target based on mlModel.

Specified by:

predict in class ACoreMLAlgorithm

Parameters:

source - Cache

target - Cache

mlModel - result of training phase

Returns:

the mapping from source to target

init

public void init(List<LearningParameter> lp,
                 ACache sourceCache,
                 ACache targetCache)

Description copied from class: ACoreMLAlgorithm

Initialize the core ML algorithm.

Overrides:

init in class ACoreMLAlgorithm

Parameters:

lp - learning parameters

sourceCache - the source cache

targetCache - the target cache

setDefaultParameters

public void setDefaultParameters()

Description copied from class: ACoreMLAlgorithm

Set default ACoreMLAlgorithm parameters values

Specified by:

setDefaultParameters in class ACoreMLAlgorithm

learn

protected MLResults learn(PseudoFMeasure pfm)
                   throws UnsupportedMLImplementationException

Description copied from class: ACoreMLAlgorithm

Learning method for unsupervised core ML algorithm implementations.

Specified by:

learn in class ACoreMLAlgorithm

Parameters:

pfm - pseudo F-measure for unsupervised learning

Returns:

wrap with results

Throws:

UnsupportedMLImplementationException - if ML implementation is not supported

supports

protected boolean supports(MLImplementationType mlType)

Description copied from class: ACoreMLAlgorithm

Check whether the mlType is supported.

Specified by:

supports in class ACoreMLAlgorithm

Parameters:

mlType - machine learning implementation type

Returns:

a boolean value

learn

protected MLResults learn(AMapping trainingData)
                   throws UnsupportedMLImplementationException

Description copied from class: ACoreMLAlgorithm

Learning method for supervised core ML algorithm implementations, where the confidence values for each pair in the trainingData determine its truth degree.

Specified by:

learn in class ACoreMLAlgorithm

Parameters:

trainingData - used for learning

Returns:

wrap with results

Throws:

UnsupportedMLImplementationException - if ML implementation is not supported

getSourceCache

public ACache getSourceCache()

Overrides:

getSourceCache in class ACoreMLAlgorithm

Returns:

the source cache

getTargetCache

public ACache getTargetCache()

Overrides:

getTargetCache in class ACoreMLAlgorithm

Returns:

the target cache

setSourceCache

public void setSourceCache(ACache sourceCache)

setTargetCache

public void setTargetCache(ACache targetCache)

getConfiguration

public Configuration getConfiguration()

Overrides:

getConfiguration in class ACoreMLAlgorithm

Returns:

the configuration

setConfiguration

public void setConfiguration(Configuration configuration)

Overrides:

setConfiguration in class ACoreMLAlgorithm

Parameters:

configuration - the configuration

getMlresult

public MLResults getMlresult()

setInitialMapping

public void setInitialMapping(AMapping initialMapping)

getTestSourceCache

public ACache getTestSourceCache()

getTestTargetCache

public ACache getTestTargetCache()

getNextExamples

protected AMapping getNextExamples(int size)
                            throws UnsupportedMLImplementationException

Description copied from class: ACoreMLAlgorithm

Get a set of examples to be added to the mapping.

Specified by:

getNextExamples in class ACoreMLAlgorithm

Parameters:

size - of the examples

Returns:

the mapping

Throws:

UnsupportedMLImplementationException - Exception