QuasiNewton Class Reference

An optimizer using the Quasi Newton method, also known as the variable metrics method. More...

Collaboration diagram for QuasiNewton:

[legend]

List of all members.

Public Member Functions
const Matrix	data ()
const Matrix	data ()
index_t	dimension ()
index_t	dimension ()
void	Eval (double *pt)
void	Eval (double *pt)
void	Init (long double(*fun)(Vector &, const Matrix &, Vector *), Matrix &data, datanode *opt_module)
void	Init (long double(*fun)(Vector &, const Matrix &, Vector *), Matrix &data, datanode *opt_module)
void	LineSearch_ (Vector pold, long double fold, Vector *grad, Vector *xi, Vector *pnew, long double *f_min, long double maximum_step_length)
void	LineSearch_ (Vector pold, long double fold, Vector *grad, Vector *xi, Vector *pnew, long double *f_min, long double maximum_step_length)

---

Detailed Description

An optimizer using the Quasi Newton method, also known as the variable metrics method.

It does multivariate minimization of an objective function using only the function value and the gradients.

Example use:

 double init_pt[d];
 index_t number_of_iters;
 struct datanode *opt_module = fx_submodule(NULL,"QuasiNewton","opt_module");
 Matrix data;
 index_t dim_param_space;

 ...
 QuasiNewton opt;
 opt.Init(obj_function, data, dim_param_space, opt_module);
 ...
 opt.Eval(init_pt);
 // init_pt contains the optimal point found

Definition at line 182 of file optimizers.h.

---

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

• fastlib/optimization/contrib/optimizers.h
• include/fastlib/optimization/contrib/optimizers.h
• fastlib/optimization/contrib/optimizers.cc