A class implementing the BFGS iterations on the. More...

#include <grad_descend.h>

Inheritance diagram for Grad_Descend:

Public Member Functions
	Grad_Descend (void(f_pointer)(Matrix_real, void , double , Matrix_real &), void meta_data_in)
	Constructor of the class. More...

	Grad_Descend (void(f_pointer)(Matrix_real, void , double , Matrix_real &), void(export_pointer)(double, Matrix_real &, void ), void meta_data_in)
	Constructor of the class. More...

double	Start_Optimization (Matrix_real &x, long maximal_iterations_in=5001)
	Call this method to start the optimization. More...

	~Grad_Descend ()
	Destructor of the class. More...

Protected Member Functions
void	get_f_ang_gradient (Matrix_real &x, double &f, Matrix_real &g)
	Call this method to obtain the cost function and its gradient at a gives position given by x. More...

void	get_Maximal_Line_Search_Step (Matrix_real &search_direction, double &maximal_step, double &search_direction__grad_overlap)
	Call this method to obtain the maximal step size during the line search. More...

virtual void	get_search_direction (Matrix_real &g, Matrix_real &search_direction, double &search_direction__grad_overlap)
	Method to get the search direction in the next line search. More...

void	line_search (Matrix_real &x, Matrix_real &g, Matrix_real &search_direction, Matrix_real &x0_search, Matrix_real &g0_search, double &maximal_step, double &d__dot__g, double &f)
	Call to perform inexact line search terminated with Wolfe 1st and 2nd conditions. More...

virtual void	Optimize (Matrix_real &x, double &f)
	Call this method to start the optimization process. More...

Protected Attributes
void(*	costfnc__and__gradient )(Matrix_real x, void params, double f, Matrix_real &g)
	function pointer to evaluate the cost function and its gradient vector More...

void(*	export_fnc )(double, Matrix_real &, void *)
	function pointer to evaluate the cost function and its gradient vector More...

long	function_call_count
	number of function calls during the optimization process More...

long	maximal_iterations
	maximal count of iterations during the optimization More...

void *	meta_data
	additional data needed to evaluate the cost function More...

double	num_precision
	numerical precision used in the calculations More...

enum solver_status	status
	status of the solver More...

int	variable_num
	number of independent variables in the problem More...

Detailed Description

A class implementing the BFGS iterations on the.

Definition at line 31 of file grad_descend.h.

Constructor & Destructor Documentation

◆ Grad_Descend() [1/2]

Grad_Descend::Grad_Descend	(	void()(Matrix_real, void , double *, Matrix_real &)	f_pointer,
		void *	meta_data_in
	)

Constructor of the class.

Parameters

f_pointer	A function pointer (x, meta_data, f, grad) to evaluate the cost function and its gradients. The cost function and the gradient vector are returned via reference by the two last arguments.
meta_data	void pointer to additional meta data needed to evaluate the cost function.

Returns: An instance of the class

Definition at line 31 of file common/grad_descend.cpp.

◆ Grad_Descend() [2/2]

Grad_Descend::Grad_Descend	(	void()(Matrix_real, void , double *, Matrix_real &)	f_pointer,
		void()(double, Matrix_real &, void )	export_pointer,
		void *	meta_data_in
	)

Constructor of the class.

Parameters

f_pointer	A function pointer (x, meta_data, f, grad) to evaluate the cost function and its gradients. The cost function and the gradient vector are returned via reference by the two last arguments.
meta_data	void pointer to additional meta data needed to evaluate the cost function.

Returns: An instance of the class

Definition at line 57 of file common/grad_descend.cpp.

◆ ~Grad_Descend()

Grad_Descend::~Grad_Descend ( )

Destructor of the class.

Definition at line 488 of file common/grad_descend.cpp.

Member Function Documentation

◆ get_f_ang_gradient()

void Grad_Descend::get_f_ang_gradient	(	Matrix_real &	x,
		double &	f,
		Matrix_real &	g
	)

protected

Call this method to obtain the cost function and its gradient at a gives position given by x.

Parameters

x	The array of the current coordinates
f	The value of the cost function is returned via this argument.
g	The gradient of the cost function at x.

Definition at line 443 of file common/grad_descend.cpp.

Here is the caller graph for this function:

◆ get_Maximal_Line_Search_Step()

void Grad_Descend::get_Maximal_Line_Search_Step	(	Matrix_real &	search_direction,
		double &	maximal_step,
		double &	search_direction__grad_overlap
	)

protected

Call this method to obtain the maximal step size during the line search.

providing at least 2*PI periodicity, unless the search direction component in a specific direction is to small.

Parameters

search_direction	The search direction.
maximal_step	The maximal allowed step in the search direction returned via this argument.
search_direction__grad_overlap	The overlap of the gradient with the search direction to test downhill.

Providing at least 2*PI periodicity, unless the search direction component in a specific direction is to small.

Parameters

search_direction	The search direction.
maximal_step	The maximal allowed step in the search direction returned via this argument.
search_direction__grad_overlap	The overlap of the gradient with the search direction to test downhill.

Definition at line 401 of file common/grad_descend.cpp.

Here is the caller graph for this function:

◆ get_search_direction()

void Grad_Descend::get_search_direction	(	Matrix_real &	g,
		Matrix_real &	search_direction,
		double &	search_direction__grad_overlap
	)

protectedvirtual

Method to get the search direction in the next line search.

Parameters

g	The gradient at the current coordinates.
search_direction	The search direction is returned via this argument (it is -g)
search_direction__grad_overlap	The overlap of the gradient with the search direction to test downhill.

Reimplemented in BFGS_Powell.

Definition at line 461 of file common/grad_descend.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ line_search()

void Grad_Descend::line_search	(	Matrix_real &	x,
		Matrix_real &	g,
		Matrix_real &	search_direction,
		Matrix_real &	x0_search,
		Matrix_real &	g0_search,
		double &	maximal_step,
		double &	d__dot__g0,
		double &	f
	)

protected

Call to perform inexact line search terminated with Wolfe 1st and 2nd conditions.

Parameters

x	The guess for the starting point. The coordinated of the optimized cost function are returned via x.
g	The gradient at x. The updated gradient is returned via this argument.
search_direction	The search direction.
x0_search	Stores the starting point. (automatically updated with the starting x during the execution)
g0_search	Stores the starting gradient. (automatically updated with the starting x during the execution)
maximal_step	The maximal allowed step in the search direction
d__dot__g	The overlap of the gradient with the search direction to test downhill.
f	The value of the minimized cost function is returned via this argument

Definition at line 127 of file common/grad_descend.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ Optimize()

void Grad_Descend::Optimize	(	Matrix_real &	x,
		double &	f
	)

protectedvirtual

Call this method to start the optimization process.

Parameters

x	The guess for the starting point. The coordinated of the optimized cost function are returned via x.
f	The value of the minimized cost function

Reimplemented in BFGS_Powell.

Definition at line 284 of file common/grad_descend.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

◆ Start_Optimization()

double Grad_Descend::Start_Optimization	(	Matrix_real &	x,
		long	maximal_iterations_in = `5001`
	)

Call this method to start the optimization.

Parameters

x	The initial solution guess.
maximal_iterations_in	The maximal number of function+gradient evaluations. Reaching this threshold the solver returns with the current solution.

Definition at line 83 of file common/grad_descend.cpp.

Here is the call graph for this function:

Here is the caller graph for this function:

Member Data Documentation

◆ costfncandgradient

void(* Grad_Descend::costfnc__and__gradient) (Matrix_real x, void *params, double *f, Matrix_real &g)

protected

function pointer to evaluate the cost function and its gradient vector

Definition at line 50 of file grad_descend.h.

◆ export_fnc

void(* Grad_Descend::export_fnc) (double, Matrix_real &, void *)

protected

function pointer to evaluate the cost function and its gradient vector

Definition at line 53 of file grad_descend.h.

◆ function_call_count

long Grad_Descend::function_call_count

protected

number of function calls during the optimization process

Definition at line 44 of file grad_descend.h.

◆ maximal_iterations

long Grad_Descend::maximal_iterations

protected

maximal count of iterations during the optimization

Definition at line 41 of file grad_descend.h.

◆ meta_data

void* Grad_Descend::meta_data

protected

additional data needed to evaluate the cost function

Definition at line 56 of file grad_descend.h.

◆ num_precision

double Grad_Descend::num_precision

protected

numerical precision used in the calculations

Definition at line 47 of file grad_descend.h.

◆ status

enum solver_status Grad_Descend::status

protected

status of the solver

Definition at line 59 of file grad_descend.h.

◆ variable_num

int Grad_Descend::variable_num

protected

number of independent variables in the problem

Definition at line 38 of file grad_descend.h.

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

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ Grad_Descend() [1/2]

◆ Grad_Descend() [2/2]

◆ ~Grad_Descend()

Member Function Documentation

◆ get_f_ang_gradient()

◆ get_Maximal_Line_Search_Step()

◆ get_search_direction()

◆ line_search()

◆ Optimize()

◆ Start_Optimization()

Member Data Documentation

◆ costfnc__and__gradient

◆ export_fnc

◆ function_call_count

◆ maximal_iterations

◆ meta_data

◆ num_precision

◆ status

◆ variable_num

◆ costfncandgradient