Fwd: [Dynare Git] Dynare branch, master, updated. fcceec896aee61c050052155ec16e0a4604ea42c
Hi Marco, I don't understand the change below. If the penalty is very steep, as with 1e8, you create a cliff for Newton-type optimizers. objective_function_penalty_base contains the best value of the objective function so far, so a small increase to this value should be sufficient to make sure that the point is never chosen by the optimizer. If you have an example where is doesn't work with a relative penalty of 100, then we should investigate. Maybe objective_function_penalty_base is not set properly somewhere. Best Michel -------- Original Message -------- Subject: [Dynare Git] Dynare branch, master, updated. fcceec896aee61c050052155ec16e0a4604ea42c Date: Mon, 1 Oct 2012 12:04:44 +0200 (CEST) From: sebastien@cepremap.org (Sébastien Villemot) To: commit@dynare.org This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Dynare". The branch, master has been updated via fcceec896aee61c050052155ec16e0a4604ea42c (commit) via 560771bedc10089f306cc11b9fa9b8c33f43e751 (commit) via 285e40eb2c248e140bffc351e6bcdd55c1bbc4e5 (commit) from e81f9d48ac5b5dab724ec48195301d139cef85ef (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit fcceec896aee61c050052155ec16e0a4604ea42c Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:53:50 2012 +0200 reset penalty to 1.e8. diff --git a/matlab/dsge_likelihood.m b/matlab/dsge_likelihood.m index 04599ca..d022b6e 100644 --- a/matlab/dsge_likelihood.m +++ b/matlab/dsge_likelihood.m @@ -429,7 +429,7 @@ switch DynareOptions.lik_init end if isnan(dLIK), info = 45; - fval = objective_function_penalty_base + 100; + fval = objective_function_penalty_base + 1.e8; exit_flag = 0; return end commit 560771bedc10089f306cc11b9fa9b8c33f43e751 Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:50:47 2012 +0200 Trap the case when second output of objective function is empty. diff --git a/matlab/mr_hessian.m b/matlab/mr_hessian.m index 0b059d9..9fe04dd 100644 --- a/matlab/mr_hessian.m +++ b/matlab/mr_hessian.m @@ -133,7 +133,7 @@ while i<n end f1(:,i)=fx; if outer_product_gradient, - if any(isnan(ffx)) + if any(isnan(ffx)) || isempty(ffx), ff1=ones(size(ff0)).*fx/length(ff0); else ff1=ffx; @@ -143,7 +143,7 @@ while i<n [fx, ffx]=feval(func,xh1,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults); f_1(:,i)=fx; if outer_product_gradient, - if any(isnan(ffx)) + if any(isnan(ffx)) || isempty(ffx), ff_1=ones(size(ff0)).*fx/length(ff0); else ff_1=ffx; commit 285e40eb2c248e140bffc351e6bcdd55c1bbc4e5 Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:49:58 2012 +0200 Pass new options nodisplay and graph_format to options_ diff --git a/matlab/dynare_sensitivity.m b/matlab/dynare_sensitivity.m index d688725..c5c16cd 100644 --- a/matlab/dynare_sensitivity.m +++ b/matlab/dynare_sensitivity.m @@ -34,6 +34,12 @@ options_gsa = set_default_option(options_gsa,'datafile',[]); if isfield(options_gsa,'nograph'), options_.nograph=options_gsa.nograph; end +if isfield(options_gsa,'nodisplay'), + options_.nodisplay=options_gsa.nodisplay; +end +if isfield(options_gsa,'graph_format'), + options_.graph_format=options_gsa.graph_format; +end if isfield(options_gsa,'mode_file'), options_.mode_file=options_gsa.mode_file; end ----------------------------------------------------------------------- Summary of changes: matlab/dsge_likelihood.m | 2 +- matlab/dynare_sensitivity.m | 6 ++++++ matlab/mr_hessian.m | 4 ++-- 3 files changed, 9 insertions(+), 3 deletions(-) hooks/post-receive -- Dynare _______________________________________________ Commit mailing list Commit@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/commit
HI Michel, in fact this was an excess of prudential penalty. Agree to revert at least until further investigations suggest to do so. thanks Marco On 10/1/2012 12:20 PM, Michel Juillard wrote:
Hi Marco,
I don't understand the change below. If the penalty is very steep, as with 1e8, you create a cliff for Newton-type optimizers. objective_function_penalty_base contains the best value of the objective function so far, so a small increase to this value should be sufficient to make sure that the point is never chosen by the optimizer. If you have an example where is doesn't work with a relative penalty of 100, then we should investigate. Maybe objective_function_penalty_base is not set properly somewhere.
Best
Michel
-------- Original Message -------- Subject: [Dynare Git] Dynare branch, master, updated. fcceec896aee61c050052155ec16e0a4604ea42c Date: Mon, 1 Oct 2012 12:04:44 +0200 (CEST) From: sebastien@cepremap.org (Sébastien Villemot) To: commit@dynare.org
This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Dynare".
The branch, master has been updated via fcceec896aee61c050052155ec16e0a4604ea42c (commit) via 560771bedc10089f306cc11b9fa9b8c33f43e751 (commit) via 285e40eb2c248e140bffc351e6bcdd55c1bbc4e5 (commit) from e81f9d48ac5b5dab724ec48195301d139cef85ef (commit)
Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below.
- Log ----------------------------------------------------------------- commit fcceec896aee61c050052155ec16e0a4604ea42c Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:53:50 2012 +0200
reset penalty to 1.e8.
diff --git a/matlab/dsge_likelihood.m b/matlab/dsge_likelihood.m index 04599ca..d022b6e 100644 --- a/matlab/dsge_likelihood.m +++ b/matlab/dsge_likelihood.m @@ -429,7 +429,7 @@ switch DynareOptions.lik_init end if isnan(dLIK), info = 45; - fval = objective_function_penalty_base + 100; + fval = objective_function_penalty_base + 1.e8; exit_flag = 0; return end
commit 560771bedc10089f306cc11b9fa9b8c33f43e751 Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:50:47 2012 +0200
Trap the case when second output of objective function is empty.
diff --git a/matlab/mr_hessian.m b/matlab/mr_hessian.m index 0b059d9..9fe04dd 100644 --- a/matlab/mr_hessian.m +++ b/matlab/mr_hessian.m @@ -133,7 +133,7 @@ while i<n end f1(:,i)=fx; if outer_product_gradient, - if any(isnan(ffx)) + if any(isnan(ffx)) || isempty(ffx), ff1=ones(size(ff0)).*fx/length(ff0); else ff1=ffx; @@ -143,7 +143,7 @@ while i<n [fx, ffx]=feval(func,xh1,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults); f_1(:,i)=fx; if outer_product_gradient, - if any(isnan(ffx)) + if any(isnan(ffx)) || isempty(ffx), ff_1=ones(size(ff0)).*fx/length(ff0); else ff_1=ffx;
commit 285e40eb2c248e140bffc351e6bcdd55c1bbc4e5 Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:49:58 2012 +0200
Pass new options nodisplay and graph_format to options_
diff --git a/matlab/dynare_sensitivity.m b/matlab/dynare_sensitivity.m index d688725..c5c16cd 100644 --- a/matlab/dynare_sensitivity.m +++ b/matlab/dynare_sensitivity.m @@ -34,6 +34,12 @@ options_gsa = set_default_option(options_gsa,'datafile',[]); if isfield(options_gsa,'nograph'), options_.nograph=options_gsa.nograph; end +if isfield(options_gsa,'nodisplay'), + options_.nodisplay=options_gsa.nodisplay; +end +if isfield(options_gsa,'graph_format'), + options_.graph_format=options_gsa.graph_format; +end if isfield(options_gsa,'mode_file'), options_.mode_file=options_gsa.mode_file; end
-----------------------------------------------------------------------
Summary of changes: matlab/dsge_likelihood.m | 2 +- matlab/dynare_sensitivity.m | 6 ++++++ matlab/mr_hessian.m | 4 ++-- 3 files changed, 9 insertions(+), 3 deletions(-)
hooks/post-receive -- Dynare _______________________________________________ Commit mailing list Commit@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/commit
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
-- Marco Ratto, Joint Research Centre The European Commission, TP 361, 21027 ISPRA(VA), ITALY Tel: +39 0332 78 9217 Fax: +39 0332 78 5733 marco.ratto@jrc.ec.europa.eu
Thanks Marco Best Michel On 10/01/2012 02:16 PM, Marco Ratto wrote:
HI Michel,
in fact this was an excess of prudential penalty. Agree to revert at least until further investigations suggest to do so. thanks Marco
On 10/1/2012 12:20 PM, Michel Juillard wrote:
Hi Marco,
I don't understand the change below. If the penalty is very steep, as with 1e8, you create a cliff for Newton-type optimizers. objective_function_penalty_base contains the best value of the objective function so far, so a small increase to this value should be sufficient to make sure that the point is never chosen by the optimizer. If you have an example where is doesn't work with a relative penalty of 100, then we should investigate. Maybe objective_function_penalty_base is not set properly somewhere.
Best
Michel
-------- Original Message -------- Subject: [Dynare Git] Dynare branch, master, updated. fcceec896aee61c050052155ec16e0a4604ea42c Date: Mon, 1 Oct 2012 12:04:44 +0200 (CEST) From: sebastien@cepremap.org (Sébastien Villemot) To: commit@dynare.org
This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Dynare".
The branch, master has been updated via fcceec896aee61c050052155ec16e0a4604ea42c (commit) via 560771bedc10089f306cc11b9fa9b8c33f43e751 (commit) via 285e40eb2c248e140bffc351e6bcdd55c1bbc4e5 (commit) from e81f9d48ac5b5dab724ec48195301d139cef85ef (commit)
Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below.
- Log ----------------------------------------------------------------- commit fcceec896aee61c050052155ec16e0a4604ea42c Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:53:50 2012 +0200
reset penalty to 1.e8.
diff --git a/matlab/dsge_likelihood.m b/matlab/dsge_likelihood.m index 04599ca..d022b6e 100644 --- a/matlab/dsge_likelihood.m +++ b/matlab/dsge_likelihood.m @@ -429,7 +429,7 @@ switch DynareOptions.lik_init end if isnan(dLIK), info = 45; - fval = objective_function_penalty_base + 100; + fval = objective_function_penalty_base + 1.e8; exit_flag = 0; return end
commit 560771bedc10089f306cc11b9fa9b8c33f43e751 Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:50:47 2012 +0200
Trap the case when second output of objective function is empty.
diff --git a/matlab/mr_hessian.m b/matlab/mr_hessian.m index 0b059d9..9fe04dd 100644 --- a/matlab/mr_hessian.m +++ b/matlab/mr_hessian.m @@ -133,7 +133,7 @@ while i<n end f1(:,i)=fx; if outer_product_gradient, - if any(isnan(ffx)) + if any(isnan(ffx)) || isempty(ffx), ff1=ones(size(ff0)).*fx/length(ff0); else ff1=ffx; @@ -143,7 +143,7 @@ while i<n [fx, ffx]=feval(func,xh1,DynareDataset,DynareOptions,Model,EstimatedParameters,BayesInfo,DynareResults); f_1(:,i)=fx; if outer_product_gradient, - if any(isnan(ffx)) + if any(isnan(ffx)) || isempty(ffx), ff_1=ones(size(ff0)).*fx/length(ff0); else ff_1=ffx;
commit 285e40eb2c248e140bffc351e6bcdd55c1bbc4e5 Author: Marco Ratto<marco.ratto@jrc.ec.europa.eu> Date: Mon Oct 1 08:49:58 2012 +0200
Pass new options nodisplay and graph_format to options_
diff --git a/matlab/dynare_sensitivity.m b/matlab/dynare_sensitivity.m index d688725..c5c16cd 100644 --- a/matlab/dynare_sensitivity.m +++ b/matlab/dynare_sensitivity.m @@ -34,6 +34,12 @@ options_gsa = set_default_option(options_gsa,'datafile',[]); if isfield(options_gsa,'nograph'), options_.nograph=options_gsa.nograph; end +if isfield(options_gsa,'nodisplay'), + options_.nodisplay=options_gsa.nodisplay; +end +if isfield(options_gsa,'graph_format'), + options_.graph_format=options_gsa.graph_format; +end if isfield(options_gsa,'mode_file'), options_.mode_file=options_gsa.mode_file; end
-----------------------------------------------------------------------
Summary of changes: matlab/dsge_likelihood.m | 2 +- matlab/dynare_sensitivity.m | 6 ++++++ matlab/mr_hessian.m | 4 ++-- 3 files changed, 9 insertions(+), 3 deletions(-)
hooks/post-receive -- Dynare _______________________________________________ Commit mailing list Commit@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/commit
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
-- Marco Ratto, Joint Research Centre The European Commission, TP 361, 21027 ISPRA(VA), ITALY Tel: +39 0332 78 9217 Fax: +39 0332 78 5733 marco.ratto@jrc.ec.europa.eu
_______________________________________________ Dev mailing list Dev@dynare.org https://www.dynare.org/cgi-bin/mailman/listinfo/dev
participants (2)
-
Marco Ratto -
Michel Juillard