---

SvmData Class Reference
[Svm]

#include <SvmData.h>

Inheritance diagram for SvmData:

List of all members.

Public Methods
	SvmData (int hyperplanedim_in, int nobservations_in, double penalty)
	SvmData (int hyperplanedim_in, int nobservations_in, double *X, double *d, double penalty)
	~SvmData ()
virtual void	YMult (double beta, SimpleVector &y, double alpha, SimpleVector &x)
virtual void	YTransMult (double beta, SimpleVector &y, double alpha, SimpleVector &x)
virtual double	dotCategories (SimpleVector &v)
virtual double	datanorm ()
virtual void	print ()
Static Public Methods
SvmData *	textInput (char filename[], double penalty, int &iErr)
Public Attributes
double	mPenalty
int	hyperplanedim
int	nobservations
DenseGenMatrixHandle	Yt
SimpleVectorHandle	categories

---

Detailed Description

Data class for Svm.

---

Constructor & Destructor Documentation

SvmData::SvmData (  int    hyperplanedim_in, int    nobservations_in, double    penalty )

Constructor that makes object with specified dimensions Parameters: hyperplanedim_in  dimenion of Euclidean space in which each obseration resides nobservations_in  number of observations penalty  penalty parameter for violation term in objective

SvmData::SvmData (  int    hyperplanedim_in, int    nobservations_in, double *    X, double *    d, double    penalty )

Constructor that makes object with specified dimensions, using storage already allocated Parameters: hyperplanedim_in  dimension of Euclidean space in which each observation resides nobservations_in  number of observations X  array of doubles containing observation points stored one after another d  array of doubles containing list of labels for the points penalty  penalty parameter for violation term in objective

SvmData::~SvmData (    )

Destructor for data class

---

Member Function Documentation

virtual double SvmData::datanorm (    )  [virtual]

compute the norm of the problem data Implements Data.

virtual double SvmData::dotCategories (  SimpleVector &    v )  [virtual]

form inner product of v with the categories vector

virtual void SvmData::print (    )  [inline, virtual]

print the problem data Implements Data.

SvmData* SvmData::textInput (  char    filename[], double    penalty, int &    iErr )  [static]

Reads Svm data from given input file Parameters: filename  name of input file. First entry of file is an integer "nobservations" representing number of observations. Second entry "hyperplanedim" is an integer representing dimension of the Euclidean space in which each point resides. Subsequent entries consist of sequences of (hyperplanedim) doubles representing one of the observation points, following by a single double representing teh label for that point. The labels must take on two distinct values. penalty  penalty parameter for violation term in objective iErr  returns 0 if OK, otherwise indicates error if file not found or input faulty

virtual void SvmData::YMult (  double    beta, SimpleVector &    y, double    alpha, SimpleVector &    x )  [virtual]

Perform a saxpy operation with matrix Y y <- beta y + alpha Yx

virtual void SvmData::YTransMult (  double    beta, SimpleVector &    y, double    alpha, SimpleVector &    x )  [virtual]

Perform a saxpy operation with matrix Yt y <- beta y + alpha Y^T x

---

Member Data Documentation

SimpleVectorHandle SvmData::categories

vector of dimension nobservations containing the labels of the points

int SvmData::hyperplanedim

dimension of Euclidean space in which each observation resides

double SvmData::mPenalty

penalty parameter for violation term in objective

int SvmData::nobservations

number of observations

DenseGenMatrixHandle SvmData::Yt

Y is an N x (t+1) matrix in which each row consists of t entries representing a point, followed by a "1". Each row corresponding to one of the two distinct labels is then multiplied by -1. Yt stores the the transpose of Y.

---

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

• SvmData.h

---