OneClassSVM Class Reference

Meta Model for SVM training. More...

#include <Svm.h>

Inheritance diagram for OneClassSVM:
MetaSVM Model

List of all members.

Public Member Functions

 OneClassSVM (SVM *pSVM, double fractionNu)
 ~OneClassSVM ()
 Descructor.
void setParameter (unsigned int index, double value)
 overloaded version of Model::setParameter
double getNu ()
 return regularization parameter $ \nu $
bool isFeasible ()
 ensure nu is positive

Protected Attributes

double nu
 Regualrization parameter, which directly controls the fraction of misclassified training pattern (outlier).

Detailed Description

Meta Model for SVM training.

Author:
E. Diner
Date:
2008
The one class SVM is a training scheme for classification and density estimation. It defines a positive constant $ \nu $ allows to misclassify a certain fraction of the training pattern which lie closest to the origin. For more information read: Estimating the Support of a High-Dimensional Distribution, B. Schölkopf

Definition at line 771 of file Svm.h.

Constructor & Destructor Documentation

OneClassSVM::OneClassSVM ( SVM *  pSVM,
double  fractionNu 
)
Parameters:
pSVM Pointer to the SVM to be optimized.
nu fraction of the training data that can be misclassified

Definition at line 1165 of file Svm.cpp.

References setParameter().

OneClassSVM::~OneClassSVM (  ) 

Descructor.

Definition at line 1171 of file Svm.cpp.

Member Function Documentation

double OneClassSVM::getNu (  )  [inline]

return regularization parameter $ \nu $

Definition at line 786 of file Svm.h.

References nu.

Referenced by SVM_Optimizer::init().

bool OneClassSVM::isFeasible (  )  [virtual]

ensure nu is positive

Reimplemented from MetaSVM.

Definition at line 1189 of file Svm.cpp.

References MetaSVM::isFeasible(), and nu.

void OneClassSVM::setParameter ( unsigned int  index,
double  value 
) [virtual]

overloaded version of Model::setParameter

Reimplemented from MetaSVM.

Definition at line 1176 of file Svm.cpp.

References nu.

Referenced by OneClassSVM().

Member Data Documentation

double OneClassSVM::nu [protected]

Regualrization parameter, which directly controls the fraction of misclassified training pattern (outlier).

Definition at line 796 of file Svm.h.

Referenced by getNu(), isFeasible(), and setParameter().

The documentation for this class was generated from the following files: