sopt::ConjugateGradient Class Reference

Solves $Ax = b$ for $x$, given $A$ and $b$. More...

#include <conjugate_gradient.h>

Classes
struct	Diagnostic
	Values indicating how the algorithm ran. More...
struct	DiagnosticAndResult
	Values indicating how the algorithm ran and its result;. More...

Public Member Functions
	ConjugateGradient (t_uint itermax=std::numeric_limits< t_uint >::max(), t_real tolerance=1e-8)
	Creates conjugate gradient operator. More...
virtual	~ConjugateGradient ()
template<typename VECTOR , typename T1 , typename T2 >
Diagnostic	operator() (VECTOR &x, Eigen::MatrixBase< T1 > const &A, Eigen::MatrixBase< T2 > const &b) const
	Computes $x$ for $Ax=b$. More...
template<typename VECTOR , typename T1 , typename T2 >
std::enable_if< not std::is_base_of< Eigen::EigenBase< T1 >, T1 >::value, Diagnostic >::type	operator() (VECTOR &x, T1 const &A, Eigen::MatrixBase< T2 > const &b) const
	Computes $x$ for $Ax=b$. More...
template<typename T0 , typename A_TYPE >
DiagnosticAndResult< typename T0::Scalar >	operator() (A_TYPE const &A, Eigen::MatrixBase< T0 > const &b) const
	Computes $x$ for $Ax=b$. More...
t_uint	itermax () const
	Maximum number of iterations. More...
void	itermax (t_uint const &itermax)
	Sets maximum number of iterations. More...
t_real	tolerance () const
	Tolerance criteria. More...
void	tolerance (t_real const &tolerance)
	Sets tolerance criteria. More...

Detailed Description

Solves $Ax = b$ for $x$, given $A$ and $b$.

Definition at line 14 of file conjugate_gradient.h.

Constructor & Destructor Documentation

ConjugateGradient()

sopt::ConjugateGradient::ConjugateGradient ( t_uint  itermax = std::numeric_limits<t_uint>::max(), t_real  tolerance = 1e-8  )

inline

Creates conjugate gradient operator.

Parameters

[in]	itermax	Maximum number of iterations. 0 means algorithm breaks only if convergence is reached.
[in]	tolerance	Convergence criteria

Definition at line 39 of file conjugate_gradient.h.

      : tolerance_(tolerance), itermax_(itermax) {}

~ConjugateGradient()

virtual sopt::ConjugateGradient::~ConjugateGradient ( )

inlinevirtual

Definition at line 41 of file conjugate_gradient.h.

{}

Member Function Documentation

itermax() [1/2]

t_uint sopt::ConjugateGradient::itermax ( ) const

inline

Maximum number of iterations.

0 means algorithm breaks only if convergence is reached.

Definition at line 75 of file conjugate_gradient.h.

{ return itermax_; }

itermax() [2/2]

void sopt::ConjugateGradient::itermax ( t_uint const &  itermax )

inline

Sets maximum number of iterations.

0 means algorithm breaks only if convergence is reached.

Definition at line 78 of file conjugate_gradient.h.

{ itermax_ = itermax; }

References itermax().

Referenced by itermax().

operator()() [1/3]

template<typename T0 , typename A_TYPE >

DiagnosticAndResult<typename T0::Scalar> sopt::ConjugateGradient::operator() ( A_TYPE const &  A, Eigen::MatrixBase< T0 > const &  b  ) const

inline

Computes $x$ for $Ax=b$.

Specialisation where x is constructed during call and returned. And x is a matrix rather than an array.

Definition at line 65 of file conjugate_gradient.h.

                                                                                          {
    DiagnosticAndResult<typename T0::Scalar> result;
    result.result = Vector<typename T0::Scalar>::Zero(b.size());
    *static_cast<Diagnostic *>(&result) = operator()(result.result, A, b);
    return result;
  }

References b, and sopt::ConjugateGradient::DiagnosticAndResult< T >::result.

operator()() [2/3]

template<typename VECTOR , typename T1 , typename T2 >

Diagnostic sopt::ConjugateGradient::operator() ( VECTOR &  x, Eigen::MatrixBase< T1 > const &  A, Eigen::MatrixBase< T2 > const &  b  ) const

inline

Computes $x$ for $Ax=b$.

Specialization that converts A from a matrix to a functor. This convertion is only so we write the conjugate-gradient algorithm only once for A as a matrix and A as a functor. A as a functor means A can be a complex operation, e.g. an FFT or two.

Definition at line 49 of file conjugate_gradient.h.

                                                            {
    return implementation(x, A, b);
  }

References b.

operator()() [3/3]

template<typename VECTOR , typename T1 , typename T2 >

std::enable_if<not std::is_base_of<Eigen::EigenBase<T1>, T1>::value, Diagnostic>::type sopt::ConjugateGradient::operator() ( VECTOR &  x, T1 const &  A, Eigen::MatrixBase< T2 > const &  b  ) const

inline

Computes $x$ for $Ax=b$.

Specialisation where A is a functor and b and x are matrix-like objects. This is the innermost specialization.

Definition at line 58 of file conjugate_gradient.h.

                                                                         {
    return implementation(x, details::wrap<typename VECTOR::PlainObject>(A), b);
  }

References b.

tolerance() [1/2]

t_real sopt::ConjugateGradient::tolerance ( ) const

inline

Tolerance criteria.

Definition at line 80 of file conjugate_gradient.h.

{ return tolerance_; }

Referenced by main(), and tolerance().

tolerance() [2/2]

void sopt::ConjugateGradient::tolerance ( t_real const &  tolerance )

inline

Sets tolerance criteria.

Definition at line 82 of file conjugate_gradient.h.

                                          {
    if (tolerance <= 0e0) throw std::domain_error("Incorrect tolerance input");
    tolerance_ = tolerance;
  }

References tolerance().

---

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

• /home/runner/work/sopt/sopt/cpp/sopt/conjugate_gradient.h