pt.tumba.ngram.svm
Class Kernel

java.lang.Object
  pt.tumba.ngram.svm.Kernel

Direct Known Subclasses:

ONECLASSQ, SVCQ, SVRQ

public abstract class Kernel
extends java.lang.Object

Abstract interface to model an SVM Kernel.

Author:

Bruno Martins

Field Summary

private double	coef0
private double	degree
private double	gamma
private int	kernelType SVM Kernel type
private SVMNode[][]	x
private double[]	xSquare

Constructor Summary

Kernel(int l, SVMNode[][] x_, SVMParameter param)
Kernel constructor, prepares to calculate the l*l kernel matrix.

Method Summary

(package private) static double	dot(SVMNode[] x, SVMNode[] y) Return the dot product of two vectors.
(package private) abstract float[]	getQ(int column, int len) Return one column of the Kernel matrix.
(package private) double	kernelFunction(int i, int j) Returns the value of a single kernel evaluation.
(package private) static double	kernelFunction(SVMNode[] x, SVMNode[] y, SVMParameter param) Returns the value of a single kernel evaluation.
(package private) void	swapIndex(int i, int j) Swap two nodes of the Kernel matrix.
private static double	tanh(double x) Returns the hyperbolic tangent of a given value

Methods inherited from class java.lang.Object

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

Field Detail

x

private SVMNode[][] x

xSquare

private final double[] xSquare

kernelType

private final int kernelType

SVM Kernel type

degree

private final double degree

gamma

private final double gamma

coef0

private final double coef0

Constructor Detail

Kernel

Kernel(int l,
       SVMNode[][] x_,
       SVMParameter param)

Kernel constructor, prepares to calculate the l*l kernel matrix.

Parameters:

l - The length of the kernel matrix (number of rows and columns).

x_ - The SVM Nodes.

param - Kernel parameters.

Method Detail

getQ

abstract float[] getQ(int column,
                      int len)

Return one column of the Kernel matrix.

Parameters:

column - The column of the kernel matrix.

len - The nu

Returns:

One column of the Kernel matrix.

swapIndex

void swapIndex(int i,
               int j)

Swap two nodes of the Kernel matrix.

Parameters:

i -

j -

tanh

private static double tanh(double x)

Returns the hyperbolic tangent of a given value

Parameters:

x - A double value.

Returns:

The hyperbolic tangent of the given value.

dot

static double dot(SVMNode[] x,
                  SVMNode[] y)

Return the dot product of two vectors.

Parameters:

x - A vector.

y - Another vector.

Returns:

The dot product of both vectors.

kernelFunction

double kernelFunction(int i,
                      int j)

Returns the value of a single kernel evaluation.

Parameters:

i - The row of the kernel matrix.

j - The column of the kernel matrix.

Returns:

The value for the single kernel evaluation.

kernelFunction

static double kernelFunction(SVMNode[] x,
                             SVMNode[] y,
                             SVMParameter param)

Returns the value of a single kernel evaluation.

Parameters:

x - The row of the kernel matrix.

y - The column of the kernel matrix.

param - SVM kernel parameters

Returns:

The value for the single kernel evaluation.