global numf numg numH

x = struct('p',ones(4000,1));
nparams = struct('maxit',1000,'toler',1.0e-4,'delbar',100,'eta',0.1,'initdel',1,'method','chol','hessian','exact');

fprintf('objg: Newton, Problem size: '); fprintf('%d\n',length(x.p));
% tic, [inform,xnew] = Newton(@objg,x,nparams); toc,
if inform.status == 0
  fprintf('CONVERGENCE FAILURE: %d steps were taken without\n', inform.iter);
  fprintf('gradient size decreasing below %8.4g.\n\n', nparams.toler);
else
  fprintf('Success: %d steps taken\n\n', inform.iter);
end
fprintf('Ending value: '); fprintf('%8.4g\n',xnew.f);
fprintf('No. function evaluations: %d\n',numf);
fprintf('No. gradient evaluations %d\n',numg);
fprintf('Norm of ending gradient: %8.4g\n', norm(xnew.g));
fprintf('No. Hessian evaluations %d\n\n',numH);

fprintf('objg: DogLeg, Problem size: '); fprintf('%d\n',length(x.p));
% tic, [inform,xnewo] = DogLeg(@objg,x,nparams); toc,
if inform.status == 0
  fprintf('CONVERGENCE FAILURE: %d steps were taken without\n', inform.iter);
  fprintf('gradient size decreasing below %8.4g.\n\n', nparams.toler);
else
  fprintf('Success: %d steps taken\n\n', inform.iter);
end
fprintf('Ending value: '); fprintf('%8.4g\n',xnew.f);
fprintf('No. function evaluations: %d\n',numf);
fprintf('No. gradient evaluations %d\n',numg);
fprintf('Norm of ending gradient: %8.4g\n', norm(xnew.g));
fprintf('No. Hessian evaluations %d\n\n',numH);

nparams.m = 3;
fprintf('\n\nLBFGS memory parameter = %d\n',nparams.m);
fprintf('objg: LBFGS, Problem size: '); fprintf('%d\n',length(x.p));
tic, [inform,xnew] = LBFGS(@objg,x,nparams); toc,
if inform.status == 0
  fprintf('CONVERGENCE FAILURE: %d steps were taken without\n', inform.iter);
  fprintf('gradient size decreasing below %8.4g.\n\n', nparams.toler);
else
  fprintf('Success: %d steps taken\n\n', inform.iter);
end
fprintf('Ending value: %8.4g\n',xnew.f);
fprintf('No. function evaluations: %d\n',numf);
fprintf('No. gradient evaluations %d\n',numg);
fprintf('Norm of ending gradient: %8.4g\n', norm(xnew.g));
fprintf('No. Hessian evaluations %d\n\n',numH);

fprintf('objg: cgTrust, Problem size: '); fprintf('%d\n',length(x.p));
tic, [inform,xnew] = cgTrust(@objg,x,nparams); toc,
if inform.status == 0
  fprintf('CONVERGENCE FAILURE: %d steps were taken without\n', inform.iter);
  fprintf('gradient size decreasing below %8.4g.\n\n', nparams.toler);
else
  fprintf('Success: %d steps taken\n\n', inform.iter);
end
fprintf('Ending value: '); fprintf('%8.4g\n',xnew.f);
fprintf('No. function evaluations: %d\n',numf);
fprintf('No. gradient evaluations %d\n',numg);
fprintf('Norm of ending gradient: %8.4g\n', norm(xnew.g));
fprintf('No. Hessian evaluations %d\n',numH);
fprintf('No. cg iterations %d\n\n',inform.cgits);