algo::nm::LevenbergMarquardt Class Reference

This class solves following equation. $$ | f(x) |^{2} = 0 $$. More...

#include <LevenbergMarquardt.h>

Inheritance diagram for algo::nm::LevenbergMarquardt:

Public Member Functions
	LevenbergMarquardt (const typename INewtonMethod::function_type &f, const typename INewtonMethod::derivative_type &derivative, const std::size_t maxIteration, const double tolerance, const std::shared_ptr< IDumpingFactorCalculator > &dumpingFactorCalculator)

Private Member Functions
virtual boost::numeric::ublas::vector< double >	doSolve (const boost::numeric::ublas::vector< double > &initialValue) const override
boost::numeric::ublas::vector< double >	calculateUpdateVector (const boost::numeric::ublas::vector< double > &x, boost::numeric::ublas::matrix< double > &inverseMatrix) const
void	calculateInverseMatrix (const boost::numeric::ublas::vector< double > &residual, const boost::numeric::ublas::matrix< double > &jacobianMatrix, boost::numeric::ublas::matrix< double > &inverseMatrix) const

Private Attributes
INewtonMethod::function_type	_f
INewtonMethod::derivative_type	_derivative
const std::size_t	_maxIteration
const double	_tolerance
std::shared_ptr< IDumpingFactorCalculator >	_dumpingFactorCalculator

Friends
class	nm_test::LevenbergMarquardtTest

Detailed Description

This class solves following equation. $$ | f(x) |^{2} = 0 $$.

Levenberg Marquardt method updates $x$ by the following equation each iteration. $$ x = x - (J_{f}^{T}(x)J_{f}(x) + {diag}(J_{f}^{T}(x)J_{f}(x)))^{-1}J_{f}^{T}(x) f(x) $$

Constructor & Destructor Documentation

algo::nm::LevenbergMarquardt::LevenbergMarquardt	(	const typename INewtonMethod::function_type &	f,
		const typename INewtonMethod::derivative_type &	derivative,
		const std::size_t	maxIteration,
		const double	tolerance,
		const std::shared_ptr< IDumpingFactorCalculator > &	dumpingFactorCalculator
	)

Parameters

f

$f,

Member Function Documentation

void algo::nm::LevenbergMarquardt::calculateInverseMatrix ( const boost::numeric::ublas::vector< double > &  residual, const boost::numeric::ublas::matrix< double > &  jacobianMatrix, boost::numeric::ublas::matrix< double > &  inverseMatrix  ) const

private

Parameters

residual
jacobianMatrix
inverseMatrix

ublas::vector< double > algo::nm::LevenbergMarquardt::calculateUpdateVector ( const boost::numeric::ublas::vector< double > &  x, boost::numeric::ublas::matrix< double > &  inverseMatrix  ) const

private

Parameters

inverseMatrix

cache to save memory allocaiton cost.

Returns

ublas::vector< double > algo::nm::LevenbergMarquardt::doSolve ( const boost::numeric::ublas::vector< double > &  initialValue ) const

overrideprivatevirtual

Parameters

initialValue

initial guess of this algorithm

Returns

$x$ such that $| f(x) | = 0$

---

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

• algo/nm/LevenbergMarquardt.h
• algo/nm/LevenbergMarquardt.cpp