Re: [Sasview-users] [Sasview-help] bug report for version WIN 4.1.2
Glad that your first issue was solved and sorry about not getting back to your question from the user list. Actually for particular feature requests the help list I think would be the best. also you can try emailing the developers list at developers@...642... There is currently a long list of tasks from major bug fixes, to major revisions and including feature requests which can be viewed at trac.sasview.org under tickets. A direct link to one common way to look at those would be http://trac.sasview.org/report/3 Please feel free to suggest things that you think would be the highest impact for your work Regarding "soft constraints" could you elaborate what you mean? Currently fairly complex equality constraints are supported though there are questions about how well those work when polydispersity is used. Inequality constraints are not supported though there is interest in seeing that. Doing inequality constraints correctly is much harder which is the current bottleneck. CHeers Paul On Sat, Sep 2, 2017 at 1:06 PM, Marcelo Ceolin <mrceolin@...6...> wrote:
Dear Paul. Thank you very much for your care and answer. It is true the difference between 4.x and 3.x about the default fitting status of the parameters. In fact it is clear from the tutorial, that only ready to fit parameters can be constrained. For me is enough...I should only pay more attention !!!!. I also sent to the user forum a comment/question about the way to suggest the developers the incorporation of other features to the program. In fact, the point is the possibility to introduce soft boundaries and soft constrains for the parameters and more complex constrains and relationships between them. If you think there is a possibility to discuss the point, I will be glad to give more details. Thank you in advance. Marcelo
El 2 sept. 2017 1:53 PM, "Paul Butler" <butlerpd@...4...> escribió:
Hi Marcelo,
Sorry for the delayed response and thank you for your feedback. Addressing your two issues
1) After investigating this is not so much a bug it turns out as failure to make usage clear so thank you for bringing to our attention. Basically the drop-down will only show parameters of the model that were selected to fit (i.e. checked in the appropriate fitting panel). I have tested 4.1.2 on windows 10 and it does seem to work properly. However the default when loading a model is for all parameters to be unchecked so unless you deliberately check at least one parameter (in model 1 in your case) you will see nothing in the drop-down (of M1 in your case). If you go back and check at least one parameter for fitting (probably FitPage 1 in your case?) you should then see that parameter in the dropdown.
I'm thinking a statement in the fit constraints section indicating this to be the case would be appropriate?
2) This is a "known behavior" that some (many?) of us had forgotten so thank you for reminding us. I will make sure it is made a high priority ticket. However, this is only true for the LM optimizer. if you chose any of the other four optimizers for the fitting, those limits should be respected. Please let us know if that is not the case.
Thanks again for the feedback and hope this helps a little
Cheers
Paul
On Thu, Aug 31, 2017 at 1:48 PM, Marcelo Ceolin <mrceolin@...6...> wrote:
Dear Developers of SASVIEW
I would like to report two bugs I found using the WINDOWS 4.1.2 release of SASVIEW 1) during the setup of SIMULTANEOUS fit and after selecting files to be fitted and activated the ADD CONSTRINS mode, there is no PARAMETER to select in the pull-down menu right from the parameter (please, find attached a screen capture)
[image: Imágenes integradas 1]
2) during a normal SINGLE FILE fit using the GUINIER POROD model (I guess it has nothing to do with the model), the fiting procedure do not take care of the limits imposed to the parameters.
Thank you very much for your support.
Sincerely yours.
MARCELO CEOLIN
-- *********************************************************************** Prof. Dr. Marcelo Ceolin Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas UNLP-CONICET Diagonal 113 y 64 Casilla de Correo 16 (sucursal 4) (B1906ZAA) La Plata Argentina
Tel: +54-221-4257291 <+54%20221%20425-7291> or 4257430 (int:173 or 162) Fax: +54-221-4254642 <+54%20221%20425-4642> e-mail: mceolin@...639... (work) or mrceolin@...6... (private) WEB: softmatter.quimica.unlp.edu.ar <https://softmatter.quimica.unlp.edu.ar/?page_id=5823> (main) or www.inifta.unlp.edu.ar <http://www.inifta.unlp.edu.ar/quimica_de_la_materia_blanda> ORCID ID: ORCID ID 0000-0002-2385-7803 <https://orcid.org/0000-0002-2385-7803> Google Scholar: Google Scholar <https://scholar.google.com.ar/citations?user=7LTN_j4AAAAJ&hl=es> ***********************************************************************
------------------------------------------------------------ ------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Sasview-help mailing list Sasview-help@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sasview-help
Dear Paul, Sorry about the delay answering your question about "soft constrains". I am thinking in two different but related lines. The whole idea came from the use of the program IFFEFIT developed by Scott Calvin and coworkers used to analyze EXAFS data. The first line is related to the a-priori knowledge about the value of a given parameter (just for example, R0). Some times we know that our parameter should be around a given value, say R0. In order to avoid running the fit with a fix value (that could be wrong because of lack of information), it could be possible introducing a pennalty at the level of the CHI2 parameter adding something like A*abs(R-R0) to the calculation of CHI2, being A the magnitud of the pennalty. The minimization algorithm should force R to be nearby R0 (in order to reduce CHI2) but allowing some freedom of the other parameters to avoid a crash in the fit. In his book XAFS for Everyone Scott Calvin cites the paper from Ellis and Freeman, 1995, DOI:10.1107/S0909049595006789 as the relevant reference. A second interesting line, also related to a-priori knowledge of the system, is to introduce complex relationships between parameters (let say R1=Tan(4*P1)/P2) in order to fit not the parameter R but P1 and P2. Here P1 and P2 may be or may be not parameters already present in the model (for example, you assume that R=a*T+b, being T the temperature of a given experiment and a and b new empirical parameters). Some times it is easier to relate or to constrain some parameters other than those defined in the fitting functions. Of course I understand that any development is a huge ammount of work but as a user of SASVIEW I feel me obliged to try to help improving and already very useful tool. Yours. Marcelo <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> Libre de virus. www.avg.com <http://www.avg.com/email-signature?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail> <#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2> 2017-09-02 22:50 GMT-03:00 Paul Butler <butlerpd@...4...>:
Glad that your first issue was solved and sorry about not getting back to your question from the user list. Actually for particular feature requests the help list I think would be the best. also you can try emailing the developers list at developers@...642...
There is currently a long list of tasks from major bug fixes, to major revisions and including feature requests which can be viewed at trac.sasview.org under tickets. A direct link to one common way to look at those would be http://trac.sasview.org/report/3
Please feel free to suggest things that you think would be the highest impact for your work
Regarding "soft constraints" could you elaborate what you mean? Currently fairly complex equality constraints are supported though there are questions about how well those work when polydispersity is used. Inequality constraints are not supported though there is interest in seeing that. Doing inequality constraints correctly is much harder which is the current bottleneck.
CHeers
Paul
On Sat, Sep 2, 2017 at 1:06 PM, Marcelo Ceolin <mrceolin@...6...> wrote:
Dear Paul. Thank you very much for your care and answer. It is true the difference between 4.x and 3.x about the default fitting status of the parameters. In fact it is clear from the tutorial, that only ready to fit parameters can be constrained. For me is enough...I should only pay more attention !!!!. I also sent to the user forum a comment/question about the way to suggest the developers the incorporation of other features to the program. In fact, the point is the possibility to introduce soft boundaries and soft constrains for the parameters and more complex constrains and relationships between them. If you think there is a possibility to discuss the point, I will be glad to give more details. Thank you in advance. Marcelo
El 2 sept. 2017 1:53 PM, "Paul Butler" <butlerpd@...4...> escribió:
Hi Marcelo,
Sorry for the delayed response and thank you for your feedback. Addressing your two issues
1) After investigating this is not so much a bug it turns out as failure to make usage clear so thank you for bringing to our attention. Basically the drop-down will only show parameters of the model that were selected to fit (i.e. checked in the appropriate fitting panel). I have tested 4.1.2 on windows 10 and it does seem to work properly. However the default when loading a model is for all parameters to be unchecked so unless you deliberately check at least one parameter (in model 1 in your case) you will see nothing in the drop-down (of M1 in your case). If you go back and check at least one parameter for fitting (probably FitPage 1 in your case?) you should then see that parameter in the dropdown.
I'm thinking a statement in the fit constraints section indicating this to be the case would be appropriate?
2) This is a "known behavior" that some (many?) of us had forgotten so thank you for reminding us. I will make sure it is made a high priority ticket. However, this is only true for the LM optimizer. if you chose any of the other four optimizers for the fitting, those limits should be respected. Please let us know if that is not the case.
Thanks again for the feedback and hope this helps a little
Cheers
Paul
On Thu, Aug 31, 2017 at 1:48 PM, Marcelo Ceolin <mrceolin@...6...> wrote:
Dear Developers of SASVIEW
I would like to report two bugs I found using the WINDOWS 4.1.2 release of SASVIEW 1) during the setup of SIMULTANEOUS fit and after selecting files to be fitted and activated the ADD CONSTRINS mode, there is no PARAMETER to select in the pull-down menu right from the parameter (please, find attached a screen capture)
[image: Imágenes integradas 1]
2) during a normal SINGLE FILE fit using the GUINIER POROD model (I guess it has nothing to do with the model), the fiting procedure do not take care of the limits imposed to the parameters.
Thank you very much for your support.
Sincerely yours.
MARCELO CEOLIN
-- *********************************************************************** Prof. Dr. Marcelo Ceolin Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas UNLP-CONICET Diagonal 113 y 64 Casilla de Correo 16 (sucursal 4) (B1906ZAA) La Plata Argentina
Tel: +54-221-4257291 <+54%20221%20425-7291> or 4257430 (int:173 or 162) Fax: +54-221-4254642 <+54%20221%20425-4642> e-mail: mceolin@...639... (work) or mrceolin@...6... (private) WEB: softmatter.quimica.unlp.edu.ar <https://softmatter.quimica.unlp.edu.ar/?page_id=5823> (main) or www.inifta.unlp.edu.ar <http://www.inifta.unlp.edu.ar/quimica_de_la_materia_blanda> ORCID ID: ORCID ID 0000-0002-2385-7803 <https://orcid.org/0000-0002-2385-7803> Google Scholar: Google Scholar <https://scholar.google.com.ar/citations?user=7LTN_j4AAAAJ&hl=es> ***********************************************************************
------------------------------------------------------------ ------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Sasview-help mailing list Sasview-help@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sasview-help
-- *********************************************************************** Prof. Dr. Marcelo Ceolin Instituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas UNLP-CONICET Diagonal 113 y 64 Casilla de Correo 16 (sucursal 4) (B1906ZAA) La Plata Argentina Tel: +54-221-4257291 or 4257430 (int:173 or 162) Fax: +54-221-4254642 e-mail: mceolin@...639... (work) or mrceolin@...6... (private) WEB: softmatter.quimica.unlp.edu.ar <https://softmatter.quimica.unlp.edu.ar/?page_id=5823> (main) or www.inifta.unlp.edu.ar <http://www.inifta.unlp.edu.ar/quimica_de_la_materia_blanda> ORCID ID: ORCID ID 0000-0002-2385-7803 <https://orcid.org/0000-0002-2385-7803> Google Scholar: Google Scholar <https://scholar.google.com.ar/citations?user=7LTN_j4AAAAJ&hl=es> ***********************************************************************
participants (2)
-
Marcelo Ceolin -
Paul Butler