jsat.regression

Class KernelRLS

java.lang.Object

jsat.regression.KernelRLS

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, Parameterized, Regressor, UpdateableRegressor

public class KernelRLS
extends java.lang.Object
implements UpdateableRegressor, Parameterized

Provides an implementation of the Kernel Recursive Least Squares algorithm. This algorithm updates the model one per data point, and induces sparsity by projecting data points down onto a set of basis vectors learned from the data stream.

See: Engel, Y., Mannor, S.,&Meir, R. (2004). The Kernel Recursive Least-Squares Algorithm. IEEE Transactions on Signal Processing, 52(8), 2275–2285. doi:10.1109/TSP.2004.830985

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
KernelRLS(KernelTrick k, double errorTolerance) Creates a new Kernel RLS learner

Method Summary

Modifier and Type	Method and Description
KernelRLS	clone()
void	finalizeModel() Finalizes the model.
double	getErrorTolerance() Returns the projection approximation tolerance
int	getModelSize() Returns the number of basis vectors that make up the model
double	regress(DataPoint data)
void	setErrorTolerance(double v) Sets the tolerance for errors in approximating a data point by projecting it onto the set of basis vectors.
void	setUp(CategoricalData[] categoricalAttributes, int numericAttributes) Prepares the classifier to begin learning from its UpdateableRegressor.update(jsat.classifiers.DataPoint, double) method.
boolean	supportsWeightedData()
void	train(RegressionDataSet dataSet)
void	train(RegressionDataSet dataSet, boolean parallel)
void	update(DataPoint dataPoint, double y_t) Updates the classifier by giving it a new data point to learn from.

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

KernelRLS

public KernelRLS(KernelTrick k,
                 double errorTolerance)

Creates a new Kernel RLS learner

Parameters:

k - the kernel trick to use

errorTolerance - the tolerance for errors in the projection

Method Detail

setErrorTolerance

public void setErrorTolerance(double v)

Sets the tolerance for errors in approximating a data point by projecting it onto the set of basis vectors. In general: as the tolerance increases the sparsity of the model increases but the accuracy may go down.
Values in the range 10^x ∀ x ∈ {-1, -2, -3, -4} often work well for this algorithm.

Parameters:

v - the approximation tolerance

getErrorTolerance

public double getErrorTolerance()

Returns the projection approximation tolerance

Returns:

the projection approximation tolerance

getModelSize

public int getModelSize()

Returns the number of basis vectors that make up the model

Returns:

the number of basis vectors that make up the model

finalizeModel

public void finalizeModel()

Finalizes the model. During online training, the the gram matrix and its inverse must be stored to perform updates, at the cost of O(n²) memory. One training is completed, these matrices are no longer needed - and can be removed to reclaim memory by finalizing the model. Once finalized, the model can no longer be updated - unless reset (destroying the model) by calling setUp(jsat.classifiers.CategoricalData[], int)

regress

public double regress(DataPoint data)

Specified by:

regress in interface Regressor

train

public void train(RegressionDataSet dataSet,
                  boolean parallel)

Specified by:

train in interface Regressor

train

public void train(RegressionDataSet dataSet)

Specified by:

train in interface Regressor

supportsWeightedData

public boolean supportsWeightedData()

Specified by:

supportsWeightedData in interface Regressor

clone

public KernelRLS clone()

Specified by:

clone in interface Regressor

Specified by:

clone in interface UpdateableRegressor

Overrides:

clone in class java.lang.Object

setUp

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

Description copied from interface: UpdateableRegressor

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

Specified by:

setUp in interface UpdateableRegressor

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

update

public void update(DataPoint dataPoint,
                   double y_t)

Description copied from interface: UpdateableRegressor

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

Specified by:

update in interface UpdateableRegressor

Parameters:

dataPoint - the data point to learn

y_t - the target value of the data point