Wrapper for GSL's multidimensional minimizer, when you have gradient. More...

#include <gsl_multimin.h>

Public Member Functions
	Gsl_Multimin_fdf (const gsl_multimin_fdfminimizer_type type, gsl_multimin_function_fdf fdf, const gsl_vector *x0, double step_size, double tol, bool verbosity=true)
	ctor builds the minimizer by allocating and setting up params More...

int	iterate ()
	Perform one iteration of the minimizer. More...

gsl_vector *	argmin () const
	query the current best argmin of the minimizer More...

double	min () const
	query the current best min value of the function to be minimized More...

gsl_vector *	gradient () const
	query the gradient of the function at the current location More...

const char *	name () const
	access a string describing the algorithm More...

int	test_gradient (double epsabs) const
	check the magnitude the gradient of the function now More...

void	restart ()
	Restart the minimizer at the current argmin value. More...

void	swap (Gsl_Multimin_fdf &that)
	Swap the contents of this minimizer and another (fast). More...

bool	verbosity (bool v)
	Alter the verbosity attribute of the object. More...

Detailed Description

Wrapper for GSL's multidimensional minimizer, when you have gradient.

GSL is the GNU Scientific Library.

This object is not copyable or assignable, but it does have a swap() method.

Constructor & Destructor Documentation

kjb::Gsl_Multimin_fdf::Gsl_Multimin_fdf	(	const gsl_multimin_fdfminimizer_type *	type,
		gsl_multimin_function_fdf *	fdf,
		const gsl_vector *	x0,
		double	step_size,
		double	tol,
		bool	verbosity = `true`
	)

inline

ctor builds the minimizer by allocating and setting up params

Parameters

type	indicates the minimization algorithm you would like to use; supported choices are listed at the link below: http://www.gnu.org/software/gsl/manual/html_node/Multimin-Algorithms-with-Derivatives.html
x0	Initial vector location in domain from which search begins.
step_size	please see the GSL documentation, or use clarivoyance, to determine what this should be set to; apparently it controls the size of (only) the first step.
tol	please see the GSL documentation for this one too.
verbosity	If true, then a bad iterate() call will emit a message to stderr.

Exceptions

KJB_error if the input is bad: e.g., x0 equal to NULL or cannot allocate memory.

Member Function Documentation

gsl_vector* kjb::Gsl_Multimin_fdf::argmin ( ) const

inline

query the current best argmin of the minimizer

gsl_vector* kjb::Gsl_Multimin_fdf::gradient ( ) const

inline

query the gradient of the function at the current location

int kjb::Gsl_Multimin_fdf::iterate ( )

inline

Perform one iteration of the minimizer.

Returns: GSL_SUCCESS if successful, or a GSL error code otherwise: http://www.gnu.org/software/gsl/manual/html_node/Error-Codes.html

If you want to throw an exception in the case of a failed iteration, consider using the GSL_ETX macro.

double kjb::Gsl_Multimin_fdf::min ( ) const

inline

query the current best min value of the function to be minimized

const char* kjb::Gsl_Multimin_fdf::name ( ) const

inline

access a string describing the algorithm

void kjb::Gsl_Multimin_fdf::restart ( )

inline

Restart the minimizer at the current argmin value.

Exceptions

KJB_error if an error occurs (not sure what that would be)

Bad news: If you want to restart the whole thing at a new point in the domain, I'm afraid you must instantiate a new object. If you alter argmin() (which you CAN do) then subsequent calls to iterate() will tell you that you are not making progress towards the solution.

Hint: the swap() method helps you resume work after instantiating a new object for the above purpose.

void kjb::Gsl_Multimin_fdf::swap ( Gsl_Multimin_fdf & that )

inline

Swap the contents of this minimizer and another (fast).

This method is useful if, for example, you want to restart the minimizer at a new location; just instantiate a new minimizer object at the new location and swap() it with the old minimizer object.

int kjb::Gsl_Multimin_fdf::test_gradient ( double epsabs ) const

inline

check the magnitude the gradient of the function now

Returns: GSL_SUCCESS if gradient magnitude is less than epsabs, or GSL_CONTINUE otherwise.

bool kjb::Gsl_Multimin_fdf::verbosity ( bool v )

inline

Alter the verbosity attribute of the object.

Returns: the previous value of the verbosity attribute. Currently if the object has true-verbosity then you will get an error message if iterate() fails.

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

/work/kobus/src/lib/gsl_cpp/gsl_multimin.h

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation