BFGS Class Reference

BFGS is a class including the BFGS minimization algorithm. More...

#include <BFGS.hpp>

Classes
struct	BFGS_function_struct
	Minimisation function struct. More...

Public Types
enum	OPTM_STATUS { GRADIENTWARNING = 0 , SUCCESS }
	Optimization status. More...
typedef Eigen::VectorXd	vector
	Vector type used in the minimazation algorithm.
typedef Eigen::MatrixXd	matrix
	MAtrix type used in the minimazation algorithm.

Static Public Member Functions
static int	optimizer (vector &startguess, BFGS_function_struct function, double tolerance, unsigned int iterationLimit, double initialStepsize=1.0, double c1=1e-4, double c2=0.9, double alphamax=1.0)
	Minimize a function using the BFGS algorithm. More...

Detailed Description

BFGS is a class including the BFGS minimization algorithm.

The BFGS minimization algorithm has been implemented as described in "Numerical Optimization - by Jorge Nocedal and Stephen J. Wright" chapter 6+3.

See also

../example/bfgsApp/bfgsExample.cpp

Member Enumeration Documentation

OPTM_STATUS

enum OPTM_STATUS

Optimization status.

Enumerator
GRADIENTWARNING	Indicating a problem with the numerical precision when evaluating the gradient.
SUCCESS	Indicating a successfully minimization.

Member Function Documentation

optimizer()

static int optimizer ( vector &  startguess, BFGS_function_struct  function, double  tolerance, unsigned int  iterationLimit, double  initialStepsize = 1.0, double  c1 = 1e-4, double  c2 = 0.9, double  alphamax = 1.0  )

static

Minimize a function using the BFGS algorithm.

Parameters

startguess	Start guess for the minimizer parameters. Replaced with minima solution at end of minimization.
function	BFGS_function_struct including pointers to the minimization function f, df and a void pointer to other data for the minimization function.
tolerance	Indicating when an acceptable minima has been found by evaluating if \( tolerance>||\Delta f(x)||_2 \).
iterationLimit	Maximum number of iterations for the BFGS algorithm.
initialStepsize	Initial step size for the BFGS algorithm.
c1	Value used to ensure the "strong Wolfe conditions" are satisfied with the value c1. See "Numerical Optimization - by Jorge Nocedal and Stephen J. Wright" chapter 3. Typical value = 1e-4.
c2	Value used to ensure the "strong Wolfe conditions" are satisfied with the value c2. See "Numerical Optimization - by Jorge Nocedal and Stephen J. Wright" chapter 3. Typical value = 0.9.
alphamax	Maximum stepsize used in iterations. Typical value of 1.0 is used to produce superlinear convergence of the overall algorithm.

Returns

GRADIENTWARNING on numerically precision problems SUCCESS when a minima is found.

Examples

bfgsExample.cpp.

---

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

• BFGS.hpp