it.uniroma2.sag.kelp.learningalgorithm.classification.hmm

Class SequenceClassificationLearningAlgorithm

java.lang.Object

it.uniroma2.sag.kelp.learningalgorithm.classification.hmm.SequenceClassificationLearningAlgorithm

All Implemented Interfaces:

LearningAlgorithm, MetaLearningAlgorithm

Direct Known Subclasses:

SequenceClassificationKernelBasedLearningAlgorithm, SequenceClassificationLinearLearningAlgorithm

public abstract class SequenceClassificationLearningAlgorithm
extends Object
implements LearningAlgorithm, MetaLearningAlgorithm

This class implements a sequential labeling paradigm.
Given sequences of items (each implemented as an Example and associated to one Label) this class allow to apply a generic LearningAlgorithm to use the "history" of each item in the sequence in order to improve the classification quality. In other words, the classification of each example does not depend only its representation, but it also depend on its "history", in terms of the classed assigned to the preceding examples.

This algorithms was inspired by the work of:
Y. Altun, I. Tsochantaridis, and T. Hofmann. Hidden Markov support vector machines. In Proceedings of the Twentieth International Conference on Machine Learning, 2003.

Author:

Danilo Croce

Constructor Summary

Constructors

Constructor and Description
SequenceClassificationLearningAlgorithm()

Method Summary

Methods

Modifier and Type	Method and Description
LearningAlgorithm	getBaseLearningAlgorithm()
int	getBeamSize()
List<Label>	getLabels() Returns the labels representing the concept to be learned.
int	getMaxEmissionCandidates()
SequencePredictionFunction	getPredictionFunction() Returns the predictionFunction learned during the training process
SequenceExampleGenerator	getSequenceExampleGenerator()
int	getTransitionsOrder()
void	learn(Dataset dataset) It starts the training process exploiting the provided dataset
void	reset() Resets all the learning process, returning to the default state.
void	setBaseLearningAlgorithm(LearningAlgorithm baseLearningAlgorithm)
void	setBeamSize(int beamSize)
void	setLabels(List<Label> labels) Sets the labels representing the concept to be learned.
void	setMaxEmissionCandidates(int maxEmissionCandidates)
void	setSequenceExampleGenerator(SequenceExampleGenerator sequenceExampleGenerator)

Methods inherited from class java.lang.Object

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

Methods inherited from interface it.uniroma2.sag.kelp.learningalgorithm.MetaLearningAlgorithm

getBaseAlgorithm, setBaseAlgorithm

Methods inherited from interface it.uniroma2.sag.kelp.learningalgorithm.LearningAlgorithm

duplicate

Constructor Detail

SequenceClassificationLearningAlgorithm

public SequenceClassificationLearningAlgorithm()

Method Detail

getBaseLearningAlgorithm

public LearningAlgorithm getBaseLearningAlgorithm()

Returns:

the learning algorithm devoted to the acquisition of a model after that each example has been enriched with its "history"

getBeamSize

public int getBeamSize()

Returns:

The size of the beam to be used in the decoding process. This number determines the number of possible sequences produced in the labeling process. It will also increase the process complexity.

getLabels

public List<Label> getLabels()

Description copied from interface: LearningAlgorithm

Returns the labels representing the concept to be learned.

Specified by:

getLabels in interface LearningAlgorithm

Returns:

the labels representing the concept to be learned

getMaxEmissionCandidates

public int getMaxEmissionCandidates()

Returns:

During the labeling process, each item is classified with respect to the target classes. To reduce the complexity of the labeling process, the returned variable determines the number of classes that received the highest classification scores to be considered after the classification step in the Viterbi Decoding.

getPredictionFunction

public SequencePredictionFunction getPredictionFunction()

Description copied from interface: LearningAlgorithm

Returns the predictionFunction learned during the training process

Specified by:

getPredictionFunction in interface LearningAlgorithm

Returns:

the predictionFunction learned during the training process

getSequenceExampleGenerator

public SequenceExampleGenerator getSequenceExampleGenerator()

Returns:

the class that generates examples enriched with information derived from their "history"

getTransitionsOrder

public int getTransitionsOrder()

Returns:

the number n of elements (in the sequence) whose labels are to be considered to enrich a targeted element

learn

public void learn(Dataset dataset)

Description copied from interface: LearningAlgorithm

It starts the training process exploiting the provided dataset

Specified by:

learn in interface LearningAlgorithm

Parameters:

dataset - the training data

reset

public void reset()

Description copied from interface: LearningAlgorithm

Resets all the learning process, returning to the default state. All the learned process is forgotten

Specified by:

reset in interface LearningAlgorithm

setBaseLearningAlgorithm

public void setBaseLearningAlgorithm(LearningAlgorithm baseLearningAlgorithm)

Parameters:

baseLearningAlgorithm - the learning algorithm devoted to the acquisition of a model after that each example has been enriched with its "history"

setBeamSize

public void setBeamSize(int beamSize)

Parameters:

beamSize - The size of the beam to be used in the decoding process. This number determines the number of possible sequences produced in the labeling process. It will also increase the process complexity.
NOTE: as this parameter does not affect the learning process, it can be also assigned later to the SequencePredictionFunction object returned from the method getPredictionFunction

setLabels

public void setLabels(List<Label> labels)

Description copied from interface: LearningAlgorithm

Sets the labels representing the concept to be learned.

Specified by:

setLabels in interface LearningAlgorithm

Parameters:

labels - the labels representing the concept to be learned

setMaxEmissionCandidates

public void setMaxEmissionCandidates(int maxEmissionCandidates)

Parameters:

maxEmissionCandidates - During the labeling process, each item is classified with respect to the target classes. To reduce the complexity of the labeling process, this variable determines the number of classes that received the highest classification scores to be considered after the classification step in the Viterbi Decoding.
NOTE: as this parameter does not affect the learning process, it can be also assigned later to the SequencePredictionFunction object returned from the method getPredictionFunction

setSequenceExampleGenerator

public void setSequenceExampleGenerator(SequenceExampleGenerator sequenceExampleGenerator)

Parameters:

sequenceExampleGenerator - the class that generates examples enriched with information derived from their "history"