1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
|
function [param,sigma] = simulated_moments_estimation(dataset,options,parallel)
% Performs estimation by Simulated Moments Method.
%
% INPUTS:
% xparam [double] p*1 vector of initial values for the estimated parameters.
% dataset [ ] Structure describing the data set.
% options [ ] Structure defining options for SMM.
% parallel [ ] Structure defining the parallel mode settings (optional).
%
% OUTPUTS:
% param [double] p*1 vector of point estimates for the parameters.
% sigma [double] p*p covariance matrix of the SMM estimates.
%
% SPECIAL REQUIREMENTS
% The user has to provide a file where the moment conditions are defined.
% Copyright (C) 2010-2017 Dynare Team
%
% This file is part of Dynare.
%
% Dynare is free software: you can redistribute it and/or modify
% it under the terms of the GNU General Public License as published by
% the Free Software Foundation, either version 3 of the License, or
% (at your option) any later version.
%
% Dynare is distributed in the hope that it will be useful,
% but WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
% GNU General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with Dynare. If not, see <http://www.gnu.org/licenses/>.
global M_ options_ oo_ estim_params_
% Load the dataset.
eval(dataset.name);
dataset.data = [];
for v = 1:dataset.number_of_observed_variables
eval(['dataset.data = [ dataset.data , ' dataset.variables(v,:) ' ];'])
end
data = dataset.data(dataset.first_observation:dataset.first_observation+dataset.number_of_observations,:);
% Compute sample moments and the weighting matrix.
eval(['[sample_moments,long_run_covariance] = ' M_.fname '_moments;'])
weighting_matrix = inv(long_run_covariance);
% Initialize output.
sigma = [];
param = [];
% Set options and initial condition.
options.estimated_parameters.list = [];
xparam = [];
if ~isempty(estim_params_.var_exo)
options.estimated_variances.idx = estim_params_.var_exo(:,1);
options.estimated_parameters.list = char(M_.exo_names(options.estimated_variances.idx,:));
options.estimated_parameters.nv = rows(estim_params_.var_exo);
xparam = [xparam; estim_params_.var_exo(:,2)];
end
if ~isempty(estim_params_.param_vals)
options.estimated_parameters.idx = estim_params_.param_vals(:,1);
if isempty(options.estimated_parameters.list)
options.estimated_parameters.list = char(M_.param_names(options.estimated_parameters.idx,:));
else
options.estimated_parameters.list = char(options.estimated_parameters.list,...
M_.param_names(options.estimated_parameters.idx,:));
end
options.estimated_parameters.np = rows(estim_params_.param_vals);
xparam = [xparam; estim_params_.param_vals(:,2)];
end
options.estimated_parameters.nb = rows(options.estimated_parameters.list);
options.estimated_parameters.lower_bound = NaN(options.estimated_parameters.nb,1);
options.estimated_parameters.upper_bound = NaN(options.estimated_parameters.nb,1);
options.estimated_parameters.lower_bound = [];
options.estimated_parameters.lower_bound = [options.estimated_parameters.lower_bound; ...
estim_params_.var_exo(:,3); ...
estim_params_.param_vals(:,3) ];
options.estimated_parameters.upper_bound = [];
options.estimated_parameters.upper_bound = [options.estimated_parameters.upper_bound; ...
estim_params_.var_exo(:,4); ...
estim_params_.param_vals(:,4) ];
options.number_of_simulated_sample = 0;
for i=1:length(parallel)
options.number_of_simulated_sample = options.number_of_simulated_sample + parallel(i).number_of_jobs*parallel(i).number_of_simulations;
end
options.observed_variables_idx = dataset.observed_variables_idx;
% Set up parallel mode if needed.
if nargin>2
if ~isunix
error('The parallel version of SMM estimation is not implemented for non unix platforms!')
end
[junk,hostname] = unix('hostname --fqdn');
hostname = deblank(hostname);
master_is_running_a_job = 0;
for i=1:length(parallel)
if strcmpi(hostname,parallel(i).machine)
master_is_running_a_job = 1;
break
end
end
if ~master_is_running_a_job
error('Master has to run one job!');
end
if options.optimization_routine>0
estimated_parameters_optimization_path = [NaN;xparam];
save('optimization_path.mat','estimated_parameters_optimization_path');
end
skipline()
disp('Master talks to its slaves...')
skipline()
% Save the workspace.
save('master_variables.mat','options_','M_','oo_');
% Send the workspace to each remote computer.
skipline()
for i = 1:length(parallel)
if ~strcmpi(hostname,parallel(i).machine)
unix(['scp master_variables.mat ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
end
end
for i=1:length(parallel)
% Write a bash script file to execute matlab scripts in the background
fid = fopen('call_matlab_session.sh','w');
fprintf(fid,'#!/bin/sh\n');
fprintf(fid,'unset DISPLAY\n');
fprintf(fid,['cd ' parallel(i).folder '\n']);
fprintf(fid,['nohup ' parallel(i).matlab '/matlab -nodesktop -nodisplay -nojvm < $1 > /dev/null 2>&1\n']);
fprintf(fid,'exit');
fclose(fid);
% Set the permission for this file (has to be executable)
%fileattrib('call_matlab_session.sh','+x','u');
unix(['chmod u+x call_matlab_session.sh']);
% Send the script file on each remote computer
if ~strcmpi(hostname,parallel(i).machine)
unix(['scp call_matlab_session.sh ' , parallel(i).login , '@' , parallel(i).machine , ':~/' ]);
else
unix(['cp call_matlab_session.sh ~/call_matlab_session.sh']);
end
end
% Send the files to each remote computer.
for i = 1:length(parallel)
if ~strcmpi(hostname,parallel(i).machine)
unix(['scp ' M_.fname '_steadystate.m ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
unix(['scp ' M_.fname '_moments.m ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
unix(['scp ' M_.fname '_static.m ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
if exist([M_.fname '_dynamic.c'])
use_dll_flag = 1;
unix(['scp ' M_.fname '_dynamic.c ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
else
use_dll_flag = 0;
unix(['scp ' M_.fname '_dynamic.m ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
end
end
end
% If needed, compile dynamic model mex file on each remote computer
if ~strcmpi(hostname,parallel(i).machine) && use_dll_flag
% Write a matlab script that will trigger the compilation of the mex file.
fid = fopen('compile_model.m', 'w');
fprintf(fid,[' eval(''mex -O LDFLAGS=''''-pthread -shared -Wl,--no-undefined'''' ' M_.fname '_dynamic.c'') ']);
fprintf(fid, '\n exit');
fclose(fid);
for i = 1:length(parallel)
if ~strcmpi(hostname,parallel(i).machine)
% Send the generated matlab script to the remote computer.
unix(['scp compile_model.m ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder]);
% Compile the mex file on the remote computer.
unix(['ssh ' parallel(i).login , '@' , parallel(i).machine ' ./call_matlab_session.sh compile_model.m']);
end
end
end
% Write the matlab script files for the evaluation of the simulated moment conditions
job = 0;
for i=1:length(parallel)
for j=1:parallel(i).number_of_jobs
job = job+1;
% Create random number streams
write_job(hostname, parallel(i).machine, parallel(i).dynare, ...
options.simulated_sample_size, length(sample_moments), ...
dataset.observed_variables_idx, options.estimated_variances.idx', options.estimated_parameters.idx', options.burn_in_periods, [M_.fname '_moments'], parallel(i).number_of_simulations, ...
parallel(i).number_of_threads_per_job, job, j, options.estimated_parameters.nb, options.estimated_parameters.nv, ...
options.estimated_parameters.np);
if ~strcmpi(hostname,parallel(i).machine)
unix(['scp ' , 'job' , int2str(job) , '.m ' , parallel(i).login , '@' , parallel(i).machine , ':' parallel(i).folder ]);
end
end
end
skipline()
disp('... And slaves do as ordered.')
skipline()
if exist('intermediary_results_from_master_and_slaves','dir')
unix('rm -rf intermediary_results_from_master_and_slaves');
end
unix('mkdir intermediary_results_from_master_and_slaves');
unix('chmod -R u+x intermediary_results_from_master_and_slaves');
end
skipline()
if options.optimization_routine==1
% Set options for csminwel.
H0 = 1e-4*eye(options.estimated_parameters.nb);
ct = 1e-4;
it = 1000;
vb = 2;
% Minimization of the objective function.
if nargin==2
[fval,param,grad,hessian_csminwel,itct,fcount,retcodehat] = ...
csminwel1('smm_objective',xparam,H0,[],ct,it,2,options_.gradient_epsilon,sample_moments,weighting_matrix,options);
elseif nargin>2
[fval,param,grad,hessian_csminwel,itct,fcount,retcodehat] = ...
csminwel1('smm_objective',xparam,H0,[],ct,it,2,options_.gradient_epsilon,sample_moments,weighting_matrix,options,parallel);
end
elseif options.optimization_routine==2
optim_options = optimset('display','iter','MaxFunEvals',1000000,'MaxIter',6000,'TolFun',1e-4,'TolX',1e-4);
if isfield(options_,'optim_opt')
eval(['optim_options = optimset(optim_options,' options_.optim_opt ');']);
end
if nargin==2
[param,fval,exitflag] = fminsearch('smm_objective',xparam,optim_options,sample_moments,weighting_matrix,options);
else
[param,fval,exitflag] = fminsearch('smm_objective',xparam,optim_options,sample_moments,weighting_matrix,options,parallel);
end
elseif options.optimization_routine==0% Compute the variance of the SMM estimator
load('optimization_path.mat');
tmp = sortrows(estimated_parameters_optimization_path',1);
param = tmp(1,2:end)';
% Compute gradient of the moment function (distance between sample and simulated moments).
[F,G] = dynare_gradient('moment_function',param,options_.gradient_epsilon,sample_moments,dataset,options,parallel);
V = (1+1/options.number_of_simulated_sample)*G'*long_run_covariance*G;
[param,diag(V)]
elseif options.optimization_routine<0
T = -options.optimization_routine;% length of the simulated time series.
time_series = extended_path(oo_.steady_state,T,1);
save time_series.mat;
end
function write_job(hostname, remotename, dynare_path, sample_size, number_of_moments, observed_variables_idx, variance_idx, parameters_idx, burn_in_periods, moments_file_name, number_of_simulations,threads_per_job, slave_number, job_number,nb,nv,np)
fid = fopen(['job' int2str(slave_number) '.m'],'w');
fprintf(fid,['% Generated by ' hostname '.\n\n']);
if ( strcmpi(hostname,remotename) && (job_number>1) ) || ~strcmpi(hostname,remotename)
fprintf(fid,'load(''master_variables'');\n');
fprintf(fid,'assignin(''base'',''M_'',M_);\n');
fprintf(fid,'assignin(''base'',''oo_'',oo_);\n');
fprintf(fid,'assignin(''base'',''options_'',options_);\n\n');
end
if ( strcmpi(hostname,remotename) && (job_number>1) ) || ~strcmpi(hostname,remotename)
fprintf(fid,['addpath ' dynare_path '\n']);
fprintf(fid,['dynare_config;\n\n']);
end
fprintf(fid,['simulated_moments = zeros(' int2str(number_of_moments) ',1);\n\n']);
fprintf(fid,'load(''estimated_parameters.mat'');\n');
fprintf(fid,['M_.params([' num2str(parameters_idx) ']) = xparams(' int2str(nv) '+1:' int2str(nb) ');\n\n']);
fprintf(fid,'tmp = diag(M_.Sigma_e);')
fprintf(fid,['tmp([' num2str(variance_idx) ']) = xparams(1:' int2str(nv) ').^2;\n\n']);
fprintf(fid,'M_.Sigma_e = diag(tmp);')
fprintf(fid,['stream=RandStream(''mt19937ar'',''Seed'',' int2str(slave_number) ');\n']);
if matlab_ver_less_than('7.12')
fprintf(fid,['RandStream.setDefaultStream(stream);\n\n']);
else
fprintf(fid,['RandStream.setGlobalStream(stream);\n\n']);
end
fprintf(fid,['maxNumCompThreads(' int2str(threads_per_job) ');\n\n']);
fprintf(fid,['for s = 1:' int2str(number_of_simulations) '\n'] );
fprintf(fid,[' time_series = extended_path([],' int2str(sample_size) ',1);\n']);
fprintf(fid,[' data = time_series([' int2str(observed_variables_idx) '],' int2str(burn_in_periods) '+1:' int2str(sample_size) ');\n']);
fprintf(fid,[' eval(''tmp = ' moments_file_name '(data);'');\n']);
fprintf(fid,[' simulated_moments = simulated_moments + tmp;\n']);
fprintf(fid,['end\n\n']);
fprintf(fid,['simulated_moments = simulated_moments/' int2str(number_of_simulations) ';\n']);
fprintf(fid,['save(''simulated_moments_slave_' int2str(slave_number) '.dat'',''simulated_moments'',''-ascii'');\n']);
if ~strcmpi(hostname,remotename)
fprintf(fid,['unix(''scp simulated_moments_slave_' int2str(slave_number) '.dat ' hostname ':' pwd '/intermediary_results_from_master_and_slaves '');\n']);
fprintf(fid,['unix(''rm simulated_moments_slave_' int2str(slave_number) '.dat'');\n']);
else
fprintf(fid,['unix(''cp simulated_moments_slave_' int2str(slave_number) '.dat ' 'intermediary_results_from_master_and_slaves '');\n']);
fprintf(fid,['unix(''rm simulated_moments_slave_' int2str(slave_number) '.dat'');\n']);
end
if ((job_number>1) && strcmpi(hostname,remotename)) || ~strcmpi(hostname,remotename)
fprintf(fid,'exit');
end
fclose(fid);
|
