it.uniroma2.sag.kelp.learningalgorithm.classification.passiveaggressive

Class PassiveAggressiveClassification

java.lang.Object

it.uniroma2.sag.kelp.learningalgorithm.PassiveAggressive

it.uniroma2.sag.kelp.learningalgorithm.classification.passiveaggressive.PassiveAggressiveClassification

All Implemented Interfaces:

BinaryLearningAlgorithm, ClassificationLearningAlgorithm, LearningAlgorithm, OnlineLearningAlgorithm

Direct Known Subclasses:

KernelizedPassiveAggressiveClassification, LinearPassiveAggressiveClassification

public abstract class PassiveAggressiveClassification
extends PassiveAggressive
implements ClassificationLearningAlgorithm

Online Passive-Aggressive Learning Algorithm for classification tasks. Every time an example is misclassified it is added the the current hyperplane, with the weight that solves the passive aggressive minimization problem reference:

[CrammerJLMR2006] Koby Crammer, Ofer Dekel, Joseph Keshet, Shai Shalev-Shwartz and Yoram Singer Online Passive-Aggressive Algorithms. Journal of Machine Learning Research (2006)

The standard algorithm is modified, including the fairness extention from

[FiliceECIR2014] S. Filice, G. Castellucci, D. Croce, and R. Basili. Effective Kernelized Online Learning in Language Processing Tasks. In collection of Advances in Information Retrieval, pp. 347-358, Springer International Publishing, 2014.

Author:

Simone Filice

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	PassiveAggressiveClassification.Loss

Nested classes/interfaces inherited from class it.uniroma2.sag.kelp.learningalgorithm.PassiveAggressive

PassiveAggressive.Policy

Field Summary

Fields

Modifier and Type	Field and Description
protected BinaryClassifier	classifier
protected float	cp
protected boolean	fairness
protected PassiveAggressiveClassification.Loss	loss

Fields inherited from class it.uniroma2.sag.kelp.learningalgorithm.PassiveAggressive

c, label, policy

Constructor Summary

Constructors

Constructor and Description
PassiveAggressiveClassification()

Method Summary

Methods

Modifier and Type	Method and Description
float	getC()
float	getCn()
float	getCp()
PassiveAggressiveClassification.Loss	getLoss()
BinaryClassifier	getPredictionFunction() Returns the classifier learned during the training process
boolean	isFairness()
void	learn(Dataset dataset) It starts the training process exploiting the provided dataset
BinaryMarginClassifierOutput	learn(Example example) Applies the learning process on a single example, updating its current model
void	setC(float c)
void	setCn(float cn)
void	setCp(float cp)
void	setFairness(boolean fairness)
void	setLoss(PassiveAggressiveClassification.Loss loss)

Methods inherited from class it.uniroma2.sag.kelp.learningalgorithm.PassiveAggressive

computeWeight, getLabel, getLabels, getPolicy, reset, setLabel, setLabels, setPolicy

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.LearningAlgorithm

duplicate, getLabels, reset, setLabels

Field Detail

loss

protected PassiveAggressiveClassification.Loss loss

cp

protected float cp

fairness

protected boolean fairness

classifier

protected BinaryClassifier classifier

Constructor Detail

PassiveAggressiveClassification

public PassiveAggressiveClassification()

Method Detail

isFairness

public boolean isFairness()

Returns:

the fairness

setFairness

public void setFairness(boolean fairness)

Parameters:

fairness - the fairness to set

getCp

public float getCp()

Returns:

the aggressiveness parameter for positive examples

setCp

public void setCp(float cp)

Parameters:

cp - the aggressiveness parameter for positive examples

getCn

public float getCn()

Returns:

the aggressiveness parameter for negative examples

setCn

public void setCn(float cn)

Parameters:

cn - the aggressiveness parameter for negative examples

getC

public float getC()

Overrides:

getC in class PassiveAggressive

Returns:

the aggressiveness parameter

setC

public void setC(float c)

Overrides:

setC in class PassiveAggressive

Parameters:

c - the aggressiveness to set

getLoss

public PassiveAggressiveClassification.Loss getLoss()

Returns:

the loss function type

setLoss

public void setLoss(PassiveAggressiveClassification.Loss loss)

Parameters:

loss - the loss function type to set

getPredictionFunction

public BinaryClassifier getPredictionFunction()

Description copied from interface: ClassificationLearningAlgorithm

Returns the classifier learned during the training process

Specified by:

getPredictionFunction in interface ClassificationLearningAlgorithm

Specified by:

getPredictionFunction in interface LearningAlgorithm

Returns:

the predictionFunction learned during the training process

learn

public BinaryMarginClassifierOutput learn(Example example)

Description copied from interface: OnlineLearningAlgorithm

Applies the learning process on a single example, updating its current model

Specified by:

learn in interface OnlineLearningAlgorithm

Parameters:

example - the instance to be exploited in the learning process

Returns:

the Prediction associated to example before the updating step

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

Overrides:

learn in class PassiveAggressive

Parameters:

dataset - the training data