pt.tumba.ngram.svm
Class SVMParameter

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

public class SVMParameter
extends java.lang.Object

Constants and Parameters used used in the SVM package.

Author:

Bruno Martins

Field Summary

double	C The cost parameter, for C_SVC, EPSILON_SVR and NU_SVR.
static int	C_SVC SVM formulation type type C-Support Vector classification
double	cacheSize The Kernel cache size, in MB
double	coef0 Parameter needed for kernels of type polynomial and sigmoid (default: 0)
double	degree The degree of the polynomial (for polynomial kernel type)
double	eps The stopping criteria
static int	EPSILON_SVR SVM formulation type type Epsilon-Support Vector regresion
double	gamma The gamma parameter.
int	kernelType Parameter specifying the Kernel type
static int	LINEAR Linear Kernel type
int	nrWeight The label weights, for C_SVC
double	nu The nu parameter, for NU_SVC, ONE_CLASS, and NU_SVR.
static int	NU_SVC SVM formulation type type Nu-Support Vector classification
static int	NU_SVR SVM formulation type type Nu-Support Vector regression
static int	ONE_CLASS SVM formulation type type distribution estimation
double	p The p parameter, for EPSILON_SVR
static int	POLY Polynomial kernel type
boolean	probability Boolean flag indicating if probability estimates should be done
static int	RBF Gaussian Radial Basis Functio kernel type
boolean	shrinking Boolean flag indicating the use of the shrinking heuristics
static int	SIGMOID Sigmoid kernel type
int	svmType Parameter specifying the SVM type
double[]	weight The label weights, for C_SVC
int[]	weightLabel The label weights, for C_SVC

Constructor Summary

SVMParameter()

Method Summary

java.lang.Object

clone() Clone this object (replicate the parameters)

Methods inherited from class java.lang.Object

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

Field Detail

C_SVC

public static final int C_SVC

SVM formulation type type C-Support Vector classification

See Also:

Constant Field Values

NU_SVC

public static final int NU_SVC

SVM formulation type type Nu-Support Vector classification

See Also:

Constant Field Values

ONE_CLASS

public static final int ONE_CLASS

SVM formulation type type distribution estimation

See Also:

Constant Field Values

EPSILON_SVR

public static final int EPSILON_SVR

SVM formulation type type Epsilon-Support Vector regresion

See Also:

Constant Field Values

NU_SVR

public static final int NU_SVR

SVM formulation type type Nu-Support Vector regression

See Also:

Constant Field Values

LINEAR

public static final int LINEAR

Linear Kernel type

See Also:

Constant Field Values

POLY

public static final int POLY

Polynomial kernel type

See Also:

Constant Field Values

RBF

public static final int RBF

Gaussian Radial Basis Functio kernel type

See Also:

Constant Field Values

SIGMOID

public static final int SIGMOID

Sigmoid kernel type

See Also:

Constant Field Values

svmType

public int svmType

Parameter specifying the SVM type

kernelType

public int kernelType

Parameter specifying the Kernel type

degree

public double degree

The degree of the polynomial (for polynomial kernel type)

gamma

public double gamma

The gamma parameter. For the polynomial kernel, gamma serves as inner product coefficient in the polynomial. In the case of the RBF kernel, gamma determines the RBF width. In the sigmoid kernel, gamma serves as as inner product coefficient in the hiperbolic tangent function (default: 1/(data dimension))

coef0

public double coef0

Parameter needed for kernels of type polynomial and sigmoid (default: 0)

cacheSize

public double cacheSize

The Kernel cache size, in MB

eps

public double eps

The stopping criteria

C

public double C

The cost parameter, for C_SVC, EPSILON_SVR and NU_SVR.

nrWeight

public int nrWeight

The label weights, for C_SVC

weightLabel

public int[] weightLabel

The label weights, for C_SVC

weight

public double[] weight

The label weights, for C_SVC

nu

public double nu

The nu parameter, for NU_SVC, ONE_CLASS, and NU_SVR. It controls the number of support vectors and errors.

p

public double p

The p parameter, for EPSILON_SVR

shrinking

public boolean shrinking

Boolean flag indicating the use of the shrinking heuristics

probability

public boolean probability

Boolean flag indicating if probability estimates should be done

Constructor Detail

SVMParameter

public SVMParameter()

Method Detail

clone

public java.lang.Object clone()

Clone this object (replicate the parameters)

Overrides:

clone in class java.lang.Object

Returns:

A clone of these parameters.