%% OPT10_RUN % % Modified: % % 08 January 2008 % %--------------------------------------------------------------------- % Running the Wood Function % This is used to test globalization methods. %--------------------------------------------------------------------- fprintf('---------------------------------------------------------\n') fprintf('Running testcase_10: exact solution (1, 1, 1, 1)\n') fprintf('---------------------------------------------------------\n') fname = 'opt10_fgh'; options = []; options.verbose = 0; options.method = 'newton'; options.max_iterations = 2000; options.max_fevals = 2000; fprintf('Newton:\n') options.globalization = 'line_search'; x0 = [-3; -1; -3; -1 ]; x = entrust(fname, x0, options); fprintf('Line search produced (%10.7e,%10.7e,%10.7e,%10.7e)\n\n',... x(1),x(2),x(3),x(4)) f = opt10_fgh ( x, 'f' ); fprintf('Value of F(X) = %f\n', f ); fprintf('Newton:\n') options.globalization = 'trust_region'; x = entrust(fname, x0, options); fprintf('Trust-region produced (%10.7e,%10.7e,%10.7e,%10.7e)\n\n',... x(1),x(2),x(3),x(4)) f = opt10_fgh ( x, 'f' ); fprintf('Value of F(X) = %f\n', f ); fprintf('Secant:\n') options.method = 'secant'; options.initial_hessian = eye(4); options.globalization = 'trust_region'; x = entrust(fname, x0, options); fprintf('Trust-region produced (%10.7e,%10.7e,%10.7e,%10.7e)\n\n',... x(1),x(2),x(3),x(4)) f = opt10_fgh ( x, 'f' ); fprintf('Value of F(X) = %f\n', f ); %--------------------------------------------------------------------- % Test Gauss-Newton strategies. %--------------------------------------------------------------------- fprintf('---------------------------------------------------------\n') fprintf('Running testcase_10 as least squares problem: \n') fprintf('Exact solution (1,1,1,1)\n') fprintf('---------------------------------------------------------\n') fname = 'opt10_rj'; options = []; options.verbose = 0; options.method = 'gauss_newton'; options.step_tolerance = 1.e-15; options.globalization = 'none'; options.gradient_tolerance = 1.e-10; options.max_iterations = 2000; options.max_fevals = 2000; x0 = [-3; -1; -3; -1 ]; x = entrust(fname, x0, options); fprintf('Gauss-Newton produced (%8.5e,%8.5e,%8.5e,%8.5e)\n\n',... x(1),x(2),x(3),x(4)) [ res, jac ] = opt10_rj ( x, 'f' ); fprintf('Norm of RES(X) = %f\n', norm ( res ) );