jsat.classifiers.linear.kernelized

Class BOGD

java.lang.Object

jsat.classifiers.BaseUpdateableClassifier

jsat.classifiers.linear.kernelized.BOGD

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, BinaryScoreClassifier, Classifier, UpdateableClassifier, Parameterized

public class BOGD
extends BaseUpdateableClassifier
implements BinaryScoreClassifier, Parameterized

Bounded Online Gradient Descent (BOGD) is a kernel learning algorithm that uses a bounded number of support vectors. Once the maximum number of support vectors is reached, old vectors are dropped either in a uniform random fashion, or weighted by the kernel function and the current coefficient for the vector. The later is the default method and is referred to as BOGD++.

See: Zhao, P., Wang, J., Wu, P., Jin, R.,&Hoi, S. C. H. (2012). Fast Bounded Online Gradient Descent Algorithms for Scalable Kernel-Based Online Learning. In Proceedings of the 29th International Conference on Machine Learning (pp. 169–176). Learning; Machine Learning. Retrieved from here

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
BOGD(BOGD toCopy) Copy constructor
BOGD(KernelTrick k, int budget, double eta, double reg, double maxCoeff) Creates a new BOGD++ learner using the HingeLoss
BOGD(KernelTrick k, int budget, double eta, double reg, double maxCoeff, LossC lossC) Creates a new BOGD++ learner

Method Summary

Modifier and Type	Method and Description
CategoricalResults	classify(DataPoint data) Performs classification on the given data point.
BOGD	clone()
int	getBudget() Returns the maximum number of allowed support vectors
double	getEta() Returns the learning rate in use
KernelTrick	getKernel() Returns the kernel to use
double	getMaxCoeff() Returns the maximum allowed value for any support vector
double	getRegularization() Returns the regularization parameter used
double	getScore(DataPoint dp) Returns the numeric score for predicting a class of a given data point, where the sign of the value indicates which class the data point is predicted to belong to.
static Distribution	guessEta(DataSet d) Guesses the distribution to use for the η parameter
static Distribution	guessMaxCoeff(DataSet d) Guesses the distribution to use for the MaxCoeff parameter
static Distribution	guessRegularization(DataSet d) Guesses the distribution to use for the Regularization parameter
boolean	isUniformSampling() Returns true is uniform sampling is in use, or false if the BOGD++ sampling procedure is in use
void	setBudget(int budget) Sets the budget for support vectors
void	setEta(double eta) Sets the learning rate to use for training.
void	setKernel(KernelTrick k) Sets the kernel to use
void	setMaxCoeff(double maxCoeff) Sets the maximum allowed value for any support vector allowed.
void	setRegularization(double regularization) Sets the regularization parameter used for training.
void	setUniformSampling(boolean uniformSampling) Sets whether or not support vectors should be removed by uniform sampling or not.
void	setUp(CategoricalData[] categoricalAttributes, int numericAttributes, CategoricalData predicting) Prepares the classifier to begin learning from its UpdateableClassifier.update(jsat.classifiers.DataPoint, int) method.
boolean	supportsWeightedData() Indicates whether the model knows how to train using weighted data points.
void	update(DataPoint dataPoint, int targetClass) Updates the classifier by giving it a new data point to learn from.

Methods inherited from class jsat.classifiers.BaseUpdateableClassifier

getEpochs, setEpochs, train, train, trainEpochs

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.classifiers.Classifier

train, train

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

BOGD

public BOGD(KernelTrick k,
            int budget,
            double eta,
            double reg,
            double maxCoeff)

Creates a new BOGD++ learner using the HingeLoss

Parameters:

k - the kernel trick to use

budget - the budget for support vectors to allow

eta - the learning rate to use

reg - the regularization parameter

maxCoeff - the maximum support vector coefficient to allow

BOGD

public BOGD(KernelTrick k,
            int budget,
            double eta,
            double reg,
            double maxCoeff,
            LossC lossC)

Creates a new BOGD++ learner

Parameters:

k - the kernel trick to use

budget - the budget for support vectors to allow

eta - the learning rate to use

reg - the regularization parameter

maxCoeff - the maximum support vector coefficient to allow

lossC - the loss function to use

BOGD

public BOGD(BOGD toCopy)

Copy constructor

Parameters:

toCopy - the object to make a copy of

Method Detail

setRegularization

public void setRegularization(double regularization)

Sets the regularization parameter used for training. The original paper suggests values in the range 2^x/T² for x ∈ {-3, -2, -1, 0, 1, 2, 3} where T is the number of data instances that will be trained on

Parameters:

regularization - the positive regularization parameter to use.

getRegularization

public double getRegularization()

Returns the regularization parameter used

Returns:

the regularization parameter used

setEta

public void setEta(double eta)

Sets the learning rate to use for training. The original paper suggests values in the range 2^x for x ∈ {-3, -2, -1, 0, 1, 2, 3}

Parameters:

eta - the positive learning rate to use

getEta

public double getEta()

Returns the learning rate in use

Returns:

the learning rate in use

setMaxCoeff

public void setMaxCoeff(double maxCoeff)

Sets the maximum allowed value for any support vector allowed. The original paper suggests values in the range 2^x for x ∈ {0, 1, 2, 3, 4}

Parameters:

maxCoeff - the maximum value for any support vector

getMaxCoeff

public double getMaxCoeff()

Returns the maximum allowed value for any support vector

Returns:

the maximum allowed value for any support vector

setBudget

public void setBudget(int budget)

Sets the budget for support vectors

Parameters:

budget - the allowed budget for support vectors

getBudget

public int getBudget()

Returns the maximum number of allowed support vectors

Returns:

the maximum number of allowed support vectors

setKernel

public void setKernel(KernelTrick k)

Sets the kernel to use

Parameters:

k - the kernel to use

getKernel

public KernelTrick getKernel()

Returns the kernel to use

Returns:

the kernel to use

setUniformSampling

public void setUniformSampling(boolean uniformSampling)

Sets whether or not support vectors should be removed by uniform sampling or not. The default is false, which corresponds to BOGD++.

Parameters:

uniformSampling - true to use uniform sampling, false otherwise.

isUniformSampling

public boolean isUniformSampling()

Returns true is uniform sampling is in use, or false if the BOGD++ sampling procedure is in use

Returns:

true is uniform sampling is in use, or false if the BOGD++ sampling procedure is in use

clone

public BOGD clone()

Specified by:

clone in interface BinaryScoreClassifier

Specified by:

clone in interface Classifier

Specified by:

clone in interface UpdateableClassifier

Specified by:

clone in class BaseUpdateableClassifier

setUp

public void setUp(CategoricalData[] categoricalAttributes,
                  int numericAttributes,
                  CategoricalData predicting)

Description copied from interface: UpdateableClassifier

Prepares the classifier to begin learning from its UpdateableClassifier.update(jsat.classifiers.DataPoint, int) method.

Specified by:

setUp in interface UpdateableClassifier

Parameters:

categoricalAttributes - an array containing the categorical attributes that will be in each data point

numericAttributes - the number of numeric attributes that will be in each data point

predicting - the information for the target class that will be predicted

getScore

public double getScore(DataPoint dp)

Description copied from interface: BinaryScoreClassifier

Returns the numeric score for predicting a class of a given data point, where the sign of the value indicates which class the data point is predicted to belong to.

Specified by:

getScore in interface BinaryScoreClassifier

Parameters:

dp - the data point to predict the class label of

Returns:

the score for the given data point

update

public void update(DataPoint dataPoint,
                   int targetClass)

Description copied from interface: UpdateableClassifier

Updates the classifier by giving it a new data point to learn from.

Specified by:

update in interface UpdateableClassifier

Parameters:

dataPoint - the data point to learn

targetClass - the target class of the data point

classify

public CategoricalResults classify(DataPoint data)

Description copied from interface: Classifier

Performs classification on the given data point.

Specified by:

classify in interface Classifier

Parameters:

data - the data point to classify

Returns:

the results of the classification.

supportsWeightedData

public boolean supportsWeightedData()

Description copied from interface: Classifier

Indicates whether the model knows how to train using weighted data points. If it does, the model will train assuming the weights. The values returned by this method may change depending on the parameters set for the model.

Specified by:

supportsWeightedData in interface Classifier

Returns:

true if the model supports weighted data, false otherwise

guessRegularization

public static Distribution guessRegularization(DataSet d)

Guesses the distribution to use for the Regularization parameter

Parameters:

d - the dataset to get the guess for

Returns:

the guess for the Regularization parameter

See Also:

setRegularization(double)

guessEta

public static Distribution guessEta(DataSet d)

Guesses the distribution to use for the η parameter

Parameters:

d - the dataset to get the guess for

Returns:

the guess for the η parameter

See Also:

setEta(double)

guessMaxCoeff

public static Distribution guessMaxCoeff(DataSet d)

Guesses the distribution to use for the MaxCoeff parameter

Parameters:

d - the dataset to get the guess for

Returns:

the guess for the MaxCoeff parameter

See Also:

(double)