jsat.classifiers.svm.extended

Class CPM

java.lang.Object

jsat.classifiers.svm.extended.CPM

All Implemented Interfaces:

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

public class CPM
extends java.lang.Object
implements BinaryScoreClassifier, Classifier, Parameterized

This class implements the Convex Polytope Machine (CPM), which is an extension of the Linear SVM. It is a binary classifier that has training time proportionate to the linear case, but can obtain accuracies closer to that of a kernelized SVM.
Similar to the AMM classifier, CPM uses multiple linear hyper-planes to create a non-linear classifier. Increasing the number of hyper-planes increases training/prediction time, but also increases the amount of non-linearity the model can tolerate.

While the CPM implements the BinaryScoreClassifier interface, the CPM decision algorithm does not completely lend itself to producing a natural score in this manner. For this reason you may observe unusual behavior from the CPM if you rely on this interface, compared with other approaches.
See: Kantchelian, A., Tschantz, M. C., Huang, L., Bartlett, P. L., Joseph, A. D., & Tygar, J. D. (2014). Large-margin Convex Polytope Machine. In Proceedings of the 27th International Conference on Neural Information Processing Systems (pp. 3248–3256). Cambridge, MA, USA: MIT Press. Retrieved from here

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
CPM() Creates a new CPM classifier, with default parameters that should work well for most cases.
CPM(CPM toCopy) Copy constructor
CPM(double lambda) Creates a new CPM classifier
CPM(double lambda, int K) Creates a new CPM classifier
CPM(double lambda, int K, double entropyThreshold) Creates a new CPM classifier
CPM(double lambda, int K, double entropyThreshold, int epochs) Creates a new CPM classifier
CPM(int K) Creates a new CPM classifier

Method Summary

Modifier and Type	Method and Description
CategoricalResults	classify(DataPoint data) Performs classification on the given data point.
CPM	clone()
double	getEntropyThreshold()
int	getEpochs() Returns the number of epochs used for training
int	getK()
double	getLambda()
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	guessEntropyThreshold(DataSet d) Provides a distribution of reasonable values for the setEntropyThreshold(double) parameter
static Distribution	guessLambda(DataSet d) Provides a distribution of reasonable values for the λ parameter
void	setEntropyThreshold(double entropyThreshold) Sets the entropy threshold used for training.
void	setEpochs(int epochs) Sets the number of whole iterations through the training set that will be performed for training
void	setK(int K) Sets the number of hyper planes to use when training.
void	setLambda(double lambda) Sets the regularization parameter λ to use.
boolean	supportsWeightedData() Indicates whether the model knows how to train using weighted data points.
void	train(ClassificationDataSet dataSet, boolean parallel) Trains the classifier and constructs a model for classification using the given data set.

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.classifiers.Classifier

train

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

CPM

public CPM()

Creates a new CPM classifier, with default parameters that should work well for most cases.

CPM

public CPM(int K)

Creates a new CPM classifier

Parameters:

K - the number of hyper-planes to learn with.

CPM

public CPM(double lambda)

Creates a new CPM classifier

Parameters:

lambda - the regularization parameter

CPM

public CPM(double lambda,
           int K)

Creates a new CPM classifier

Parameters:

lambda - the regularization parameter

K - the number of hyper-planes to learn with.

CPM

public CPM(double lambda,
           int K,
           double entropyThreshold)

Creates a new CPM classifier

Parameters:

lambda - the regularization parameter

K - the number of hyper-planes to learn with.

entropyThreshold - the parameter that encourages non-linearity to be exploited

CPM

public CPM(double lambda,
           int K,
           double entropyThreshold,
           int epochs)

Creates a new CPM classifier

Parameters:

lambda - the regularization parameter

K - the number of hyper-planes to learn with.

entropyThreshold - the parameter that encourages non-linearity to be exploited

epochs - the number of training iterations

CPM

public CPM(CPM toCopy)

Copy constructor

Parameters:

toCopy - the object to copy

Method Detail

setEntropyThreshold

public void setEntropyThreshold(double entropyThreshold)

Sets the entropy threshold used for training. It ensures a diversity of hyper-planes are used, where larger values encourage using more of the hyper planes.

This method is adjusted from the paper's definition so that the input can be any non-negative value. It is recommended to try values in the range of [0, 10]

Parameters:

entropyThreshold - the non-negative parameter for hyper-plane diversity

getEntropyThreshold

public double getEntropyThreshold()

Returns:

the non-negative parameter for hyper-plane diversity

setLambda

public void setLambda(double lambda)

Sets the regularization parameter λ to use. Larger values penalize model complexity. This value is adjusted from the form in the original paper so that you do not need to consider the number of epochs explicitly. The effective regularization will be divided by the total number of training updates.

Parameters:

lambda - the regularization parameter value to use, the recommended range range is (0, 10⁴]

getLambda

public double getLambda()

Returns:

the regularization parameter value

setK

public void setK(int K)

Sets the number of hyper planes to use when training. A normal linear model is equivalent to using only 1 hyper plane. The more hyper planes used, the more modeling capacity the algorithm has, but the slower it will run.

Parameters:

K - the number of hyper planes to use.

getK

public int getK()

Returns:

the number of hyper planes to use.

setEpochs

public void setEpochs(int epochs)

Sets the number of whole iterations through the training set that will be performed for training

Parameters:

epochs - the number of whole iterations through the data set

getEpochs

public int getEpochs()

Returns the number of epochs used for training

Returns:

the number of epochs used for training

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.

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

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

train

public void train(ClassificationDataSet dataSet,
                  boolean parallel)

Description copied from interface: Classifier

Trains the classifier and constructs a model for classification using the given data set. If the training method knows how, it will used the threadPool to conduct training in parallel. This method will block until the training has completed.

Specified by:

train in interface Classifier

Parameters:

dataSet - the data set to train on

parallel - true if multiple threads should be used to train the model. false if it should be done in a single threaded manner.

clone

public CPM clone()

Specified by:

clone in interface BinaryScoreClassifier

Specified by:

clone in interface Classifier

Overrides:

clone in class java.lang.Object

guessLambda

public static Distribution guessLambda(DataSet d)

Provides a distribution of reasonable values for the λ parameter

Parameters:

d - the dataset to get the guess for

Returns:

the distribution to search this parameter from

guessEntropyThreshold

public static Distribution guessEntropyThreshold(DataSet d)

Provides a distribution of reasonable values for the setEntropyThreshold(double) parameter

Parameters:

d - the dataset to get the guess for

Returns:

the distribution to search this parameter from