Método numérico para optimización con rengiones de confianza

RegConfianza.m

function x = RegConfianza(fun, maxiter, tol, delta, DELTA, etha)
    
% ===========================================================
%             METODO DE REGIONES DE CONFIANZA
%
% AUTHOR: Jose Manuel Proudinat Silva
% ID: 000130056
% COURSE: Analisis Aplicado
%
% DESCRIPTION:
%
%
% INPUT:
%
%
% OUTPUT:
%
%
% ===========================================================


% Configuracion AMPL
nombre_ampl = strcat(fun, '.nl');
salida = strcat(fun, '.out');
fout = fopen(salida, 'w');

[x, ~, ~, ~, ~, ~] = spamfunc(nombre_ampl);
    
% Valores iniciales
n = length(x)
k = 1;
[f, ~] = spamfunc(x, 0);
[g, ~] = spamfunc(x, 1);
[H, ~] = spamfunc(x, 2);
normg = norm(g);
curv = 1;
mf = @(H, g, f, x) x' * H * x + g' * x + f;

% Preparamos la impresion
fprintf( fout, '######################################################################\n\n');
fprintf( fout, ' Nombre del problema      %10s \n', fun);
fprintf( fout, ' Numero de variables      %3i \n', n);
fprintf(fout,'   k        f            ||g||     ||p||     nfg     curv  \n');
fprintf(fout,['----------------------------------------------------' ...
              '-------------------\n']);

while k < maxiter && normg > tol

    % Obtenemos la direccion de descenso a traves del GC modificado
    p = GCTR(H, -g, tol, tol, maxiter, delta);

    % Calculamos el paso rho
    xx = x + p;
    [ff, ~] = spamfunc(xx, 0);
    mm = mf(H, g, f, p);
    rho = (f - ff) / (f - mm);

    % Actualizamos la region de confianza
    normp = norm(p);

    if rho < 0.25
        delta = 0.25 * normp;
    else
        if rho > 0.75 && normp == delta
            delta = min(2 * delta, DELTA);
        end
    end
    
    if rho > etha
        x = xx;
    end

    k = k + 1;

    if mod(k, 50) == 0
        fprintf(fout, ' %3i  %14.8e   %8.2e   %5.3f    %3i       %5.3f   \n', ...
                            k, f,  normg, normp, numfg, curv );
    end
    
end

    
end