OSKL

Documentation of 'jsat.classifiers.linear.kernelized.OSKL' Java class OSKL

All Classes

Summary:
Nested |
Field |
Constr |
Method

Detail:
Field |
Constr |
Method

java.lang.Object
- jsat.classifiers.BaseUpdateableClassifier
- - jsat.classifiers.linear.kernelized.OSKL

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, BinaryScoreClassifier, Classifier, UpdateableClassifier, Parameterized
```
public class OSKL
extends BaseUpdateableClassifier
implements BinaryScoreClassifier, Parameterized
```
Online Sparse Kernel Learning by Sampling and Smooth Losses (OSKL) is an online algorithm for learning sparse kernelized solutions to binary classification problems. The number of support vectors is controlled by a a sparsity parameter G and a specified loss function. The number of support vectors is bounded by the cumulative loss of the loss function used.

The OSKL algorithm is designed for use with smooth loss functions such as the logistic loss. However, it can work with non-smooth loss functions such as the hinge loss.

See: Zhang, L., Yi, J., Jin, R., Lin, M.,&He, X. (2013). Online Kernel Learning with a Near Optimal Sparsity Bound. In S. Dasgupta&D. Mcallester (Eds.), Proceedings of the 30th International Conference on Machine Learning (ICML-13) (Vol. 28, pp. 621–629). JMLR Workshop and Conference Proceedings.

See Also:

Serialized Form

Constructor Summary

Constructors
Constructor and Description
`OSKL(KernelTrick k, double R)` Creates a new OSKL learner using the `LogisticLoss`.
`OSKL(KernelTrick k, double eta, double G, double R)` Creates a new OSKL learner using the `LogisticLoss`
`OSKL(KernelTrick k, double eta, double G, double R, LossC lossC)` Creates a new OSKL learner
`OSKL(OSKL toCopy)` Copy constructor

Method Summary

All Methods Static Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`CategoricalResults`	`classify(DataPoint data)` Performs classification on the given data point.
`OSKL`	`clone()`
`int`	`getBurnIn()` Returns the number of burn in rounds
`double`	`getEta()` Returns the learning rate in use
`double`	`getG()` Returns the sparsification parameter
`KernelTrick`	`getKernel()` Returns the kernel to use
`double`	`getR()` Returns the maximum allowed norm for the model learned
`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.
`int`	`getSupportVectorCount()` Returns the number of data points accepted as support vectors
`static Distribution`	`guessR(DataSet d)` Guesses the distribution to use for the R parameter
`boolean`	`isUseAverageModel()` Returns `true` if the average of all models is being used, or `false` if the last model is used
`void`	`setBurnIn(int burnIn)` Sets the number of update calls to consider as part of the "burn in" phase.
`void`	`setEta(double eta)` Sets the learning rate to use for training.
`void`	`setG(double G)` Sets the sparsification parameter G.
`void`	`setKernel(KernelTrick k)` Sets the kernel to use
`void`	`setR(double R)` Sets the maximum allowed norm of the model.
`void`	`setUp(CategoricalData[] categoricalAttributes, int numericAttributes, CategoricalData predicting)` Prepares the classifier to begin learning from its `UpdateableClassifier.update(jsat.classifiers.DataPoint, int)` method.
`void`	`setUseAverageModel(boolean useAverageModel)` Sets whether or not the average of all intermediate models is used or if the most recent model is used when performing classification
`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
  - OSKL
```
public OSKL(KernelTrick k,
            double R)
```
    Creates a new OSKL learner using the LogisticLoss. The parameters setG(double) and setEta(double) are set based on the original papers suggestions to produced a less sparse model that should be more accurate
    
    Parameters:
    
    k - the kernel to use
    
    R - the maximum allowed norm for the model
  - OSKL
```
public OSKL(KernelTrick k,
            double eta,
            double G,
            double R)
```
    Creates a new OSKL learner using the LogisticLoss
    
    Parameters:
    
    k - the kernel to use
    
    eta - the learning rate to use
    
    G - the sparsification parameter
    
    R - the maximum allowed norm for the model
  - OSKL
```
public OSKL(KernelTrick k,
            double eta,
            double G,
            double R,
            LossC lossC)
```
    Creates a new OSKL learner
    
    Parameters:
    
    k - the kernel to use
    
    eta - the learning rate to use
    
    G - the sparsification parameter
    
    R - the maximum allowed norm for the model
    
    lossC - the loss function to use
  - OSKL
```
public OSKL(OSKL toCopy)
```
    Copy constructor
    
    Parameters:
    
    toCopy - the object to copy
- Method Detail
  - 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
  - setEta
```
public void setEta(double eta)
```
    Sets the learning rate to use for training. The original paper suggests setting η = 0.9/G
    
    Parameters:
    
    eta - the positive learning rate to use
  - getEta
```
public double getEta()
```
    Returns the learning rate in use
    
    Returns:
    
    the learning rate in use
  - setG
```
public void setG(double G)
```
    Sets the sparsification parameter G. Increasing G reduces the number of updates to the model, which increases sparsity but may reduce accuracy. Decreasing G increases the update rate reducing sparsity. The original paper tests values of G ∈ {1, 2, 4, 10}
    
    Parameters:
    
    G - the sparsification parameter in [1, ∞)
  - getG
```
public double getG()
```
    Returns the sparsification parameter
    
    Returns:
    
    the sparsification parameter
  - guessR
```
public static Distribution guessR(DataSet d)
```
    Guesses the distribution to use for the R parameter
    
    Parameters:
    
    d - the dataset to get the guess for
    
    Returns:
    
    the guess for the R parameter
    
    See Also:
    
    setR(double)
  - setR
```
public void setR(double R)
```
    Sets the maximum allowed norm of the model. The original paper suggests values in the range 10^x for x ∈ {0, 1, 2, 3, 4, 5}.
    
    Parameters:
    
    R - the maximum allowed norm for the model
  - getR
```
public double getR()
```
    Returns the maximum allowed norm for the model learned
    
    Returns:
    
    the maximum allowed norm for the model learned
  - setUseAverageModel
```
public void setUseAverageModel(boolean useAverageModel)
```
    Sets whether or not the average of all intermediate models is used or if the most recent model is used when performing classification
    
    Parameters:
    
    useAverageModel - true to use the average model, false to use the last model update
  - isUseAverageModel
```
public boolean isUseAverageModel()
```
    Returns true if the average of all models is being used, or false if the last model is used
    
    Returns:
    
    true if the average of all models is being used, or false if the last model is used
  - setBurnIn
```
public void setBurnIn(int burnIn)
```
    Sets the number of update calls to consider as part of the "burn in" phase. The averaging of the model will not start until after the burn in phase.
    If the classification or score is requested before the burn in phase is completed, the latest model will be used as is.
    
    Parameters:
    
    burnIn - the number of updates to ignore before averaging. Must be non negative.
  - getBurnIn
```
public int getBurnIn()
```
    Returns the number of burn in rounds
    
    Returns:
    
    the number of burn in rounds
  - 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
  - getSupportVectorCount
```
public int getSupportVectorCount()
```
    Returns the number of data points accepted as support vectors
    
    Returns:
    
    the number of support vectors in the model
  - 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
  - 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
  - clone
```
public OSKL 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

Skip navigation links

All Classes

Summary:
Nested |
Field |
Constr |
Method

Detail:
Field |
Constr |
Method

Class OSKL

Constructor Summary

Method Summary

Methods inherited from class jsat.classifiers.BaseUpdateableClassifier

Methods inherited from class java.lang.Object

Methods inherited from interface jsat.classifiers.Classifier

Methods inherited from interface jsat.parameters.Parameterized

Constructor Detail

OSKL

OSKL

OSKL

OSKL

Method Detail

setKernel

getKernel

setEta

getEta

setG

getG

guessR

setR

getR

setUseAverageModel

isUseAverageModel

setBurnIn

getBurnIn

setUp

getSupportVectorCount

update

classify

supportsWeightedData

getScore

clone