Class min_quad_golden (o2scl)

O2scl : Class List

template<class func_t = funct>
class min_quad_golden : public o2scl::min_bkt_base<funct>

Minimization of a function using the safeguarded step-length algorithm of Gill and Murray [GSL].

This class is unfinished.

Documentation from GSL:

This algorithm performs univariate minimization (i.e., line
search). It requires only objective function values g(x) to
compute the minimum. The algorithm maintains an interval of
uncertainty [a,b] and a point x in the interval [a,b] such that
a < x < b, and g(a) > g(x) and g(x) < g(b). The algorithm also
maintains the three points with the smallest objective values x,
v and w such that g(x) < g(v) < g(w). The algorithm terminates
when max( x-a, b-x ) < 2(r |x| + t) where r and t are small
positive reals. At a given iteration, the algorithm first fits a
quadratic through the three points (x, g(x)), (v, g(v)) and (w,
g(w)) and computes the location of the minimum u of the
resulting quadratic. If u is in the interval [a,b] then g(u) is
computed. If u is not in the interval [a,b], and either v < x
and w < x, or v > x and w > x (i.e., the quadratic is
extrapolating), then a point u' is computed using a safeguarding
procedure and g(u') is computed. If u is not in the interval
[a,b], and the quadratic is not extrapolating, then a point u''
is computed using approximate golden section and g(u'') is
computed. After evaluating g() at the appropriate new point, a,
b, x, v, and w are updated and the next iteration is performed.
The algorithm is based on work presented in the following
references.
                                                                      
Algorithms for Minimization without derivatives
Richard Brent
Prentice-Hall Inc., Englewood Cliffs, NJ, 1973

Safeguarded Steplength Algorithms for Optimization using Descent Methods
Philip E. Gill and Walter Murray
Division of Numerical Analysis and Computing
National Physical Laboratory, Teddington, United Kingdom
NPL Report NAC 37, August 1974

Idea for Future:

Take common elements of this and min_brent and move to a generic GSL minimizer type?

Public Functions

inline min_quad_golden()
inline virtual ~min_quad_golden()
inline int set(func_t &func, double xmin, double lower, double upper)

Set the function and the initial bracketing interval.

inline int set_with_values(func_t &func, double xmin, double fmin, double lower, double fl, double upper, double fu)

Set the function, the initial bracketing interval, and the corresponding function values.

inline int iterate()

Perform an iteration.

inline virtual int min_bkt(double &x2, double x1, double x3, double &fmin, func_t &func)

Calculate the minimum fmin of func with x2 bracketed between x1 and x3.

Note that this algorithm requires that the initial guess already brackets the minimum, i.e. \( x_1<x_2<x_3 \), \( f(x_1)>f(x_2) \) and \( f(x_3)>f(x_2) \). This is different from min_cern, where the initial value of the first parameter to min_cern::min_bkt() is ignored.

inline virtual const char *type()

Return string denoting type (“min_quad_golden”)

Public Members

double x_minimum

Location of minimum.

double x_lower

Lower bound.

double x_upper

Upper bound.

double f_minimum

Minimum value.

double f_lower

Value at lower bound.

double f_upper

Value at upper bound.

Protected Functions

inline int compute_f_values(func_t &func, double xminimum, double *fminimum, double xlower, double *flower, double xupper, double *fupper)

Compute the function values at the various points.

Protected Attributes

func_t *uf

The function.

double x_prev_small
double f_prev_small
double x_small
double f_small
double step_size
double stored_step
double prev_stored_step
size_t num_iter
double rel_err_val
double abs_err_val
double golden_mean
double golden_ratio