com.imsl.datamining.supportvectormachine

Class Kernel

java.lang.Object

com.imsl.datamining.supportvectormachine.Kernel

All Implemented Interfaces:

Cloneable

Direct Known Subclasses:

LinearKernel, PolynomialKernel, RadialBasisKernel, SigmoidKernel

public abstract class Kernel
extends Object
implements Cloneable

Abstract class to specify a kernel function for support vector machines.

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	Kernel(int numberOfParameters) Creates a Kernel and specifies the number of kernel parameters for a specific Kernel.

Method Summary

Modifier and Type	Method and Description
abstract Kernel	clone() Returns a clone of this object.
protected double	dot(DataNode[] x, DataNode[] y) Calculates the dot product between two DataNode arrays.
double[]	getParameters() Returns the kernel parameters.
abstract double	kernelFunction(DataNode[][] x, int i, int j) Abstract method to calculate the kernel function between two DataNode arrays.
abstract double	kernelFunction(DataNode[] x, DataNode[] y) Abstract method to calculate the kernel function between two DataNode arrays.
void	setParameters(double... kParams) Sets the kernel parameters.

Methods inherited from class java.lang.Object

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

Constructor Detail

Kernel

protected Kernel(int numberOfParameters)

Creates a Kernel and specifies the number of kernel parameters for a specific Kernel.

Parameters:

numberOfParameters - an int indicating the number of parameters this Kernel must have

Method Detail

kernelFunction

public abstract double kernelFunction(DataNode[] x,
                                      DataNode[] y)

Abstract method to calculate the kernel function between two DataNode arrays.

Parameters:

x - a DataNode array

y - a DataNode array

Returns:

a double, the kernel function evaluated at x and y

kernelFunction

public abstract double kernelFunction(DataNode[][] x,
                                      int i,
                                      int j)

Abstract method to calculate the kernel function between two DataNode arrays.

Parameters:

x - a DataNode matrix

i - an int, the index of the first DataNode, x[i]

j - an int, the index of the second DataNode, x[j]

Returns:

a double, the kernel function evaluated at x[i] and x[j]

clone

public abstract Kernel clone()

Returns a clone of this object.

Overrides:

clone in class Object

Returns:

a Kernel which is a clone of this object

getParameters

public double[] getParameters()

Returns the kernel parameters.

Returns:

a double array containing the kernel parameters

setParameters

public void setParameters(double... kParams)

Sets the kernel parameters.

Specific instances can override this method to test for proper values of the parameters.

Parameters:

kParams - a double array containing the parameters

dot

protected double dot(DataNode[] x,
                     DataNode[] y)

Calculates the dot product between two DataNode arrays. The array lengths may be different.

Parameters:

x - a DataNode array

y - a DataNode array

Returns:

a double, the dot product between x and y