function [x, ithist, iflag] = newtnd( f, x, tolf, tolx, maxit ) % % function [x, ithist, iflag] = newtnd( f, x, tolf, tolx, maxit ) % % fdnewtnd attempts to compute a root of F using Newton's method % % Input parameters: % f name of a matlab function that evaluates % f and its derivative. % x initial iterate % tolf stopping tolerance (optional. Default tolf = 1.e-7) % Newton's method stops if ||F(x)||_2 < tolf % tolx stopping tolerance (optional. Default tolx = 1.e-7) % Newton's method stops if ||s||_2 < tolx, % where s = -F'(x)^{-1} F(x) is the Newton step. % maxit maximum number of iterations (optional. Default maxit = 100) % % % Output parameters: % x approximation of the solution. % ithist array with the iteration history % The i-th row of ithist contains [it, x, F, s, t] % ifag return flag % iflag = 0 ||F(x)||_2 <= tolf % iflag = 1 iteration terminated because maximum number of % iterations was reached. ||F(x)||_2 > tolf % iflag = 2 iteration terminated because no step size was found % % Matthias Heinkenschloss % Department of Computational and Applied Mathematics % Rice University % March 11, 2004 % % set tolerances if necessary if( nargin <= 3 ) tolf = 1.e-7; tolx = 1.e-7; maxit = 100; end if( nargin <= 4 ) tolx = 1.e-7; maxit = 100; end if( nargin <= 5 ) maxit = 100; end it = 0; iflag = 0; [F,Jac] = feval(f, x); normF = norm(F); s = tolx*ones(size(x)); t = 1; while( it < maxit & norm(t*s) > tolx & normF > tolf ) s = - (Jac\F); t = 1; xtmp = x+s; F = feval(f, xtmp); while ( norm(F)^2 > (1-t*2.e-4)*normF^2 & norm(t*s) > tolx ) t = t/2; xtmp = x+t*s; F = feval(f, xtmp); end ithist(it+1,:) = [it, norm(x), normF, norm(t*s), t]; x = xtmp; it = it+1; [F,Jac] = feval(f, x); normF = norm(F); end % check why the Newton's method truncated and set iflag if( norm(F) > tolf ) % Newton's method truncated because the maximum number of iterations % was reached iflag = 1; return elseif( norm(t*s) > tolx & t < 1 ) iflag = 2; return else % Newton's method truncated because norm(F) <= tolf % print info for last iteration ithist(it+1,:) = [it, norm(x), norm(F),0,0]; end