Problem formulation for binary C-SVM problems. More...

#include <shark/Algorithms/QP/QpSolver.h>

Public Types
typedef MatrixT	MatrixType

typedef MatrixType::QpFloatType	QpFloatType

Public Member Functions
	CSVMProblem (MatrixType &quadratic, Data< unsigned int > const &labels, double C)
	Setup only using kernel matrix, labels and regularization parameter.

	CSVMProblem (MatrixType &quadratic, Data< unsigned int > const &labels, RealVector const &regularizers)
	Setup using kernel matrix, labels and different regularization parameters for positive and negative classes.

	CSVMProblem (MatrixType &quadratic, RealVector linear, Data< unsigned int > const &labels, double C)

std::size_t	dimensions () const

double	boxMin (std::size_t i) const

double	boxMax (std::size_t i) const

void	flipCoordinates (std::size_t i, std::size_t j)
	exchange two variables via the permutation

void	scaleBoxConstraints (double factor, double variableScalingFactor)
	Scales all box constraints by a constant factor and adapts the solution by scaling it by the same factor.

Public Attributes
MatrixType &	quadratic
	representation of the quadratic part of the objective function

RealVector	linear
	Linear part of the problem.

RealVector	alpha
	Solution candidate.

RealVector	diagonal

std::vector< std::size_t >	permutation
	permutation of the variables alpha, gradient, etc.

Detailed Description

template<class MatrixT>
class shark::CSVMProblem< MatrixT >

Problem formulation for binary C-SVM problems.

\(max_\alpha - 1/2 \alpha^T K \alpha + \alpha^Ty\) under constraints: \(l_i <= \alpha_i <= u_i\) \(\sum_i \alpha_i = 0\)

Definition at line 246 of file QpSolver.h.

Member Typedef Documentation

◆ MatrixType

template<class MatrixT >

typedef MatrixT shark::CSVMProblem< MatrixT >::MatrixType

Definition at line 248 of file QpSolver.h.

◆ QpFloatType

template<class MatrixT >

typedef MatrixType::QpFloatType shark::CSVMProblem< MatrixT >::QpFloatType

Definition at line 249 of file QpSolver.h.

Constructor & Destructor Documentation

◆ CSVMProblem() [1/3]

template<class MatrixT >

shark::CSVMProblem< MatrixT >::CSVMProblem	(	MatrixType &	quadratic,
		Data< unsigned int > const &	labels,
		double	C
	)

inline

Setup only using kernel matrix, labels and regularization parameter.

Definition at line 252 of file QpSolver.h.

References shark::CSVMProblem< MatrixT >::diagonal, shark::CSVMProblem< MatrixT >::dimensions(), shark::Data< Type >::element(), shark::CSVMProblem< MatrixT >::linear, shark::Data< Type >::numberOfElements(), shark::CSVMProblem< MatrixT >::permutation, shark::CSVMProblem< MatrixT >::quadratic, and SIZE_CHECK.

◆ CSVMProblem() [2/3]

template<class MatrixT >

shark::CSVMProblem< MatrixT >::CSVMProblem	(	MatrixType &	quadratic,
		Data< unsigned int > const &	labels,
		RealVector const &	regularizers
	)

inline

Setup using kernel matrix, labels and different regularization parameters for positive and negative classes.

Definition at line 274 of file QpSolver.h.

References shark::CSVMProblem< MatrixT >::diagonal, shark::CSVMProblem< MatrixT >::dimensions(), shark::Data< Type >::element(), shark::CSVMProblem< MatrixT >::linear, shark::Data< Type >::numberOfElements(), shark::CSVMProblem< MatrixT >::permutation, shark::CSVMProblem< MatrixT >::quadratic, and SIZE_CHECK.

◆ CSVMProblem() [3/3]

template<class MatrixT >

shark::CSVMProblem< MatrixT >::CSVMProblem	(	MatrixType &	quadratic,
		RealVector	linear,
		Data< unsigned int > const &	labels,
		double	C
	)

inline

Definition at line 302 of file QpSolver.h.

References shark::CSVMProblem< MatrixT >::diagonal, shark::CSVMProblem< MatrixT >::dimensions(), shark::Data< Type >::element(), shark::CSVMProblem< MatrixT >::linear, shark::Data< Type >::numberOfElements(), shark::CSVMProblem< MatrixT >::permutation, shark::CSVMProblem< MatrixT >::quadratic, and SIZE_CHECK.

Member Function Documentation

◆ boxMax()

template<class MatrixT >

double shark::CSVMProblem< MatrixT >::boxMax ( std::size_t i ) const

inline

Definition at line 330 of file QpSolver.h.

◆ boxMin()

template<class MatrixT >

double shark::CSVMProblem< MatrixT >::boxMin ( std::size_t i ) const

inline

Definition at line 327 of file QpSolver.h.

◆ dimensions()

template<class MatrixT >

std::size_t shark::CSVMProblem< MatrixT >::dimensions ( ) const

inline

Definition at line 323 of file QpSolver.h.

References shark::CSVMProblem< MatrixT >::quadratic.

Referenced by shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), and shark::CSVMProblem< MatrixT >::scaleBoxConstraints().

◆ flipCoordinates()

template<class MatrixT >

void shark::CSVMProblem< MatrixT >::flipCoordinates	(	std::size_t	i,
		std::size_t	j
	)

inline

exchange two variables via the permutation

Definition at line 351 of file QpSolver.h.

References shark::CSVMProblem< MatrixT >::alpha, shark::CSVMProblem< MatrixT >::diagonal, shark::CSVMProblem< MatrixT >::linear, shark::CSVMProblem< MatrixT >::permutation, and shark::CSVMProblem< MatrixT >::quadratic.

◆ scaleBoxConstraints()

template<class MatrixT >

void shark::CSVMProblem< MatrixT >::scaleBoxConstraints	(	double	factor,
		double	variableScalingFactor
	)

inline

Scales all box constraints by a constant factor and adapts the solution by scaling it by the same factor.

Definition at line 365 of file QpSolver.h.

References shark::CSVMProblem< MatrixT >::alpha, and shark::CSVMProblem< MatrixT >::dimensions().

Member Data Documentation

◆ alpha

template<class MatrixT >

RealVector shark::CSVMProblem< MatrixT >::alpha

Solution candidate.

Definition at line 341 of file QpSolver.h.

Referenced by shark::CSVMProblem< MatrixT >::flipCoordinates(), and shark::CSVMProblem< MatrixT >::scaleBoxConstraints().

◆ diagonal

template<class MatrixT >

RealVector shark::CSVMProblem< MatrixT >::diagonal

diagonal matrix entries The diagonal array is of fixed size and not subject to shrinking.

Definition at line 345 of file QpSolver.h.

Referenced by shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), and shark::CSVMProblem< MatrixT >::flipCoordinates().

◆ linear

template<class MatrixT >

RealVector shark::CSVMProblem< MatrixT >::linear

Linear part of the problem.

Definition at line 338 of file QpSolver.h.

Referenced by shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), and shark::CSVMProblem< MatrixT >::flipCoordinates().

◆ permutation

template<class MatrixT >

std::vector<std::size_t> shark::CSVMProblem< MatrixT >::permutation

permutation of the variables alpha, gradient, etc.

Definition at line 348 of file QpSolver.h.

Referenced by shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), and shark::CSVMProblem< MatrixT >::flipCoordinates().

◆ quadratic

template<class MatrixT >

MatrixType& shark::CSVMProblem< MatrixT >::quadratic

representation of the quadratic part of the objective function

Definition at line 335 of file QpSolver.h.

Referenced by shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::CSVMProblem(), shark::CSVMProblem< MatrixT >::dimensions(), and shark::CSVMProblem< MatrixT >::flipCoordinates().

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

include/shark/Algorithms/QP/QpSolver.h

Public Types

Public Member Functions

Public Attributes

Detailed Description

Member Typedef Documentation

◆ MatrixType

◆ QpFloatType

Constructor & Destructor Documentation

◆ CSVMProblem() [1/3]

◆ CSVMProblem() [2/3]

◆ CSVMProblem() [3/3]

Member Function Documentation

◆ boxMax()

◆ boxMin()

◆ dimensions()

◆ flipCoordinates()

◆ scaleBoxConstraints()

Member Data Documentation

◆ alpha

◆ diagonal

◆ linear

◆ permutation

◆ quadratic