stepan
/
dynare
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fixed issues related to newratflag. 

This options relates to alternative numerical hessian computations:

optim=('Hessian',1) is the default dynare numeric Hessian
optim=('Hessian',0) is the outer product gradient
optim=('Hessian',2) is the 'mixed' outer product gradient, where diagonal elements using second order derivation formula,

Both 0 and 2 cases require univariate filters, to ensure using maximum number of individual densities.
time-shift
Marco Ratto 2015-03-12 17:24:21 +01:00
parent 87cb00bd7d
commit c881cfff17
3 changed files with 35 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
matlab/dynare_estimation_1.m
View File

@ -233,7 +233,7 @@ end
if ~isequal(options_.mode_compute,0) && ~options_.mh_posterior_mode_estimation
%prepare settings for newrat
if options_.mode_compute==5
%get whether analytical Hessian with non-analytical mode-finding is requested
%get whether outer product Hessian is requested
newratflag=[];
if ~isempty(options_.optim_opt)
options_list = read_key_value_string(options_.optim_opt);
@ -246,19 +246,14 @@ if ~isequal(options_.mode_compute,0) && ~options_.mh_posterior_mode_estimation
if options_.analytic_derivation,
options_analytic_derivation_old = options_.analytic_derivation;
options_.analytic_derivation = -1;
if ~isempty(newratflag) && newratflag~=0 %gradient explicitly specified
if ~isempty(newratflag) && newratflag~=0 %numerical hessian explicitly specified
error('newrat: analytic_derivation is incompatible with numerical Hessian.')
else %use default
newratflag=0; %use analytical gradient
newratflag=0; %exclude DYNARE numerical hessian
end
elseif ~options_.analytic_derivation
if isempty(newratflag)
newratflag=options_.newrat.hess; %use default gradient
end
if newratflag==0 %Analytic Hessian wanted, but not automatically computed by newrat itself
if ~((options_.kalman_algo == 2) || (options_.kalman_algo == 4)) %kalman_algo not compatible
error('Analytical Hessian with non-analytical mode-finding requires kalman_algo=2 or 4.')
end
newratflag=options_.newrat.hess; %use default numerical dynare hessian
end
end
end
@ -281,11 +276,39 @@ if ~isequal(options_.mode_compute,0) && ~options_.mh_posterior_mode_estimation
[junk1, junk2, hh] = feval(objective_function,xparam1, ...
dataset_,dataset_info,options_,M_,estim_params_,bayestopt_,bounds,oo_);
options_.analytic_derivation = ana_deriv_old;
elseif ~(isequal(options_.mode_compute,5) && newratflag==0),
elseif ~(isequal(options_.mode_compute,5) && newratflag~=1),
% with flag==0, we force to use the hessian from outer
% product gradient of optimizer 5
hh = reshape(hessian(objective_function,xparam1, ...
options_.gstep,dataset_,dataset_info,options_,M_,estim_params_,bayestopt_,bounds,oo_),nx,nx);
elseif isequal(options_.mode_compute,5)
% other numerical hessian options available with optimizer 5
%
% if newratflag == 0
% compute outer product gradient of optimizer 5
%
% if newratflag == 2
% compute 'mixed' outer product gradient of optimizer 5
% with diagonal elements computed with numerical second order derivatives
%
% uses univariate filters, so to get max # of available
% densitities for outer product gradient
kalman_algo0 = options_.kalman_algo;
compute_hessian = 1;
if ~((options_.kalman_algo == 2) || (options_.kalman_algo == 4)),
options_.kalman_algo=2;
if options_.lik_init == 3,
options_.kalman_algo=4;
end
elseif newratflag==0, % hh already contains outer product gradient with univariate filter
compute_hessian = 0;
end
if compute_hessian,
crit = options_.newrat.tolerance.f;
newratflag = newratflag>0;
hh = reshape(mr_hessian(0,xparam1,objective_function,newratflag,crit,dataset_, dataset_info, options_,M_,estim_params_,bayestopt_,bounds,oo_), nx, nx);
end
options_.kalman_algo = kalman_algo0;
end
end
end

2
matlab/global_initialization.m
View File

@ -481,7 +481,7 @@ csminwel.maxiter=1000;
options_.csminwel=csminwel;
%newrat optimization routine
newrat.hess=1; %analytic hessian
newrat.hess=1; % dynare numerical hessian
newrat.tolerance.f=1e-5;
newrat.tolerance.f_analytic=1e-7;
newrat.maxiter=1000;

3
matlab/optimization/dynare_minimize_objective.m
View File

@ -177,10 +177,9 @@ switch minimizer_algorithm
end
end
[opt_par_values,hessian_mat,gg,fval,invhess] = newrat(objective_function,start_par_value,analytic_grad,crit,nit,0,varargin{:});
%hessian_mat is the plain outer product gradient Hessian
if options_.analytic_derivation %Hessian is already analytic one, reset option
options_.analytic_derivation = ana_deriv;
elseif ~options_.analytic_derivation && newratflag ==0 %Analytic Hessian wanted, but not computed yet
hessian_mat = reshape(mr_hessian(0,opt_par_values,objective_function,1,crit,varargin{:}), n_params, n_params);
end
case 6
[opt_par_values, hessian_mat, Scale, fval] = gmhmaxlik(objective_function, start_par_value, ...