---

SvmVars Class Reference
[Svm]

#include <SvmVars.h>

Inheritance diagram for SvmVars:

List of all members.

Public Methods
	SvmVars (int hyperplanedim_in, int nobservations_in)
	SvmVars (int hyperplanedim_in, int nobservations_in, double w[], double v[], double z[], double u[], double s[])
virtual	~SvmVars ()
virtual void	saxpy (Variables *b, double alpha)
virtual void	negate ()
virtual double	mu ()
virtual double	mustep (Variables *step, double alpha)
virtual double	stepbound (Variables *b)
virtual void	interiorPoint (double alpha, double beta)
virtual void	shiftBoundVariables (double alpha, double beta)
virtual double	violation ()
virtual void	print ()
virtual void	printCoefs ()
virtual void	copy (Variables *b)
virtual double	findBlocking (Variables *step, double &primalValue, double &primalStep, double &dualValue, double &dualStep, int &firstOrSecond)
virtual void	asMfile (ostream &os)
virtual double	onenorm ()
virtual double	infnorm ()
Public Attributes
int	hyperplanedim
int	nobservations
SimpleVectorHandle	z
SimpleVectorHandle	w
SimpleVectorHandle	u
SimpleVectorHandle	v
SimpleVectorHandle	s

---

Detailed Description

Variables class for Svm.

---

Constructor & Destructor Documentation

SvmVars::SvmVars (  int    hyperplanedim_in, int    nobservations_in )

constructor to make a Svm variables object of specified dimensions

SvmVars::SvmVars (  int    hyperplanedim_in, int    nobservations_in, double    w[], double    v[], double    z[], double    u[], double    s[] )

constructor to make a Svm variables object of specified dimensions for which storage has already been allocated for the five individual vector components

virtual SvmVars::~SvmVars (    )  [virtual]

destructor for Svm variables object

---

Member Function Documentation

virtual void SvmVars::asMfile (  ostream &    os )  [virtual]

write current set of variables to given output stream

virtual void SvmVars::copy (  Variables *    b )  [virtual]

copy the variables Implements Variables.

virtual double SvmVars::findBlocking (  Variables *    step, double &    primalValue, double &    primalStep, double &    dualValue, double &    dualStep, int &    firstOrSecond )  [virtual]

Performs the same function as stepbound, and supplies additional information about which component of the nonnegative variables is responsible for restricting alpha. In terms of the abstract formulation, the components have the following meanings. Parameters: primalValue  the value of the blocking component of the primal variables s. primalStep  the corresponding value of the blocking component of the primal step variables b->s. dualValue  the value of the blocking component of the dual variables z. dualStep  the corresponding value of the blocking component of the dual step variables b->z. firstOrSecond  1 if the primal step is blocking, 2 if the dual step is block, 0 if no step is blocking. See also: stepbound Implements Variables.

virtual double SvmVars::infnorm (    )  [virtual]

compute the inf-norm of the variables Implements Variables.

virtual void SvmVars::interiorPoint (  double    alpha, double    beta )  [virtual]

In the abstract QP formulation, sets s to alpha, z to beta and the other variable components to zero. Implements Variables.

virtual double SvmVars::mu (    )  [virtual]

compute complementarity gap, obtained by taking the inner product of the complementary vectors and dividing by the total number of components Implements Variables.

virtual double SvmVars::mustep (  Variables *    step, double    alpha )  [virtual]

compute the complementarity gap resulting from a step of length "alpha" along direction "step" Implements Variables.

virtual void SvmVars::negate (    )  [virtual]

negate the value of all the variables in this structure Implements Variables.

virtual double SvmVars::onenorm (    )  [virtual]

compute the 1-norm of the variables Implements Variables.

virtual void SvmVars::print (    )  [virtual]

print the variables Reimplemented from Variables.

virtual void SvmVars::printCoefs (    )  [virtual]

print only the interesting parts of the SVM variable vector (that is, the coefficient that define the separating hyperplane)

virtual void SvmVars::saxpy (  Variables *    b, double    alpha )  [virtual]

performs saxpy operation: add alpha*b to current variables Implements Variables.

virtual void SvmVars::shiftBoundVariables (  double    alpha, double    beta )  [virtual]

In the standard QP formulation, sets s += alpha, z += beta Implements Variables.

virtual double SvmVars::stepbound (  Variables *    b )  [virtual]

calculate the largest alpha in (0,1] such that the nonnegative variables stay nonnegative in the given search direction. In the abstract problem formulation, this is the largest value of alpha such that (s,z) + alpha * (b->s,b->z) >= 0. See also: findBlocking Implements Variables.

virtual double SvmVars::violation (    )  [virtual]

The amount by which the current variables violate the non-negativity constraints. Implements Variables.

---

Member Data Documentation

int SvmVars::hyperplanedim

dimension of Euclidean space in which each observation resides

int SvmVars::nobservations

number of observations

SimpleVectorHandle SvmVars::s

vectors and the scalar that make up the Svm variables; see the paper by E. M. Gertz and S. J. Wright for details of the formulation

SimpleVectorHandle SvmVars::u

vectors and the scalar that make up the Svm variables; see the paper by E. M. Gertz and S. J. Wright for details of the formulation

SimpleVectorHandle SvmVars::v

vectors and the scalar that make up the Svm variables; see the paper by E. M. Gertz and S. J. Wright for details of the formulation

SimpleVectorHandle SvmVars::w

vectors and the scalar that make up the Svm variables; see the paper by E. M. Gertz and S. J. Wright for details of the formulation

SimpleVectorHandle SvmVars::z

vectors and the scalar that make up the Svm variables; see the paper by E. M. Gertz and S. J. Wright for details of the formulation

---

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

• SvmVars.h

---