Dear all,
Do you have any idea of what means NAN (Not A Number?) in "sum path"
Ifeffit output during the fit?
I've already posted this issue in a post dated 2006/10/04 and Bruce
asked to me to give him something to reproduce this issue on his
computer but this behavior is very unstable and I don't understand how
to exactly reproduce it.
I would like to know in your opinion what could bring a variable during
the minimization process to NAN. In addition, what exactly means the
four numbers visible in the "sum path #1 #2 #3 #4" output? It seems to
me that, for each path, #2 = delr and #3 = sigma2, isn't it?
M.
--
Mauro Rovezzi
Hi Mauro, I don't know exactly what the 'sum path # # # # ' in Artemis is supposed to display, but I'd guess that some variable or fit parameter is being unstable.
I've already posted this issue in a post dated 2006/10/04 and Bruce asked to me to give him something to reproduce this issue on his computer but this behavior is very unstable and I don't understand how to exactly reproduce it.
I think we'd need to see a project that could be made to have this problem, even if it happens only occasionally.
I would like to know in your opinion what could bring a variable during the minimization process to NAN.
Divide-by-zero and log(-1) come to mind, but I think we'd need to determine whether a variable is being brought to NAN by the minimization process before speculating why. Cheers, --Matt
On Friday 20 April 2007, Mauro Rovezzi wrote:
I would like to know in your opinion what could bring a variable during the minimization process to NAN. In addition, what exactly means the four numbers visible in the "sum path #1 #2 #3 #4" output? It seems to me that, for each path, #2 = delr and #3 = sigma2, isn't it?
I don't quite understand this. Would a screenshot or a clar project file help? B -- Bruce Ravel ---------------------------------------------- bravel@anl.gov Molecular Environmental Science Group, Building 203, Room E-165 MRCAT, Sector 10, Advanced Photon Source, Building 433, Room B007 Argonne National Laboratory phone and voice mail: (1) 630 252 5033 Argonne IL 60439, USA fax: (1) 630 252 9793 My homepage: http://cars9.uchicago.edu/~ravel EXAFS software: http://cars9.uchicago.edu/~ravel/software/
Quoting Bruce Ravel
I don't quite understand this. Would a screenshot or a clar project file help?
Bruce,
I use Ifeffit and Horae packages under Debian built by Carlo so launching
Artemis from the command line:
$artemis&
at each fit, during the minimization process, appear in the standard
output, that is in the shell where I executed the program many lines (one
for each path at each minimization step, if I well understand it) that
have this form:
sum path #1 #2 #3 #4
Where #* are real numbers and sometimes appears "NAN" that, as Matt said,
seems a division by zero or a log(-1). I'm trying to produce a project
file that will show the case when the minimization fall in all "NAN" and
the loop need to be killed to stop. I'll post it as soon as possible, but
in the meantime is useful to understand these numbers that appear in the
shell.
M.
--
Mauro Rovezzi
Hi Mauro,
I use Ifeffit and Horae packages under Debian built by Carlo so launching Artemis from the command line:
If you are using a linux system, I strongly recommend compiling Ifeffit and horae from source, and indicating which Ifeffit and horae distribution you use. I cannot offer any support for the debian packages. I'm not saying the packaging is bad, or at fault here -- it is almost certainly is not. But I have no idea what is in the package, so it's pointless to try to troubleshoot it. --Matt
But I can, so please send me anything you can to indicate how I can duplicate the problem. Also, can you send me the version and distribution that you are using. Carlo On Fri, 20 Apr 2007, Matt Newville wrote:
Hi Mauro,
I use Ifeffit and Horae packages under Debian built by Carlo so launching Artemis from the command line:
If you are using a linux system, I strongly recommend compiling Ifeffit and horae from source, and indicating which Ifeffit and horae distribution you use. I cannot offer any support for the debian packages.
I'm not saying the packaging is bad, or at fault here -- it is almost certainly is not. But I have no idea what is in the package, so it's pointless to try to troubleshoot it.
--Matt _______________________________________________ Ifeffit mailing list Ifeffit@millenia.cars.aps.anl.gov http://millenia.cars.aps.anl.gov/mailman/listinfo/ifeffit
-- Carlo U. Segre -- Professor of Physics Associate Dean for Special Projects, Graduate College Illinois Institute of Technology Voice: 312.567.3498 Fax: 312.567.3494 segre@iit.edu http://www.iit.edu/~segre segre@debian.org
Hi Mauro: I wanted to close the loop on this problem. I have implemented a temporary solution while I figure out what the right thing to do is. This solution applise to whomever is using the Debian packages for etch and sid as well as my unofficial packages for sarge and the ubuntu varieties. Note that for Debian lenny you should use the sid repository. Point your /etc/apt/sources.list to whichever source applies to your distribution: deb http://debian-xray.iit.edu sarge main contrib non-free deb http://debian-xray.iit.edu etch main contrib non-free deb http://debian-xray.iit.edu sid main contrib non-free deb http://debian-xray.iit.edu edgy main contrib non-free deb http://debian-xray.iit.edu feisty main contrib non-free upgrade using whatever tools you choose. You will get a package that is names 1.3.0 but has been rolled back to the 1.2.9 version. This has been tested to not have the error that you found on etch, sid and sarge. The problem was my mistakenly believing that Matt had released version 1.3.0 in September 2006 (my fault, I should have asked!). I will keep this naming scheme until the real 1.3.0 (or 1.3.1) releases. As for the official version in Debian. i will look into what is the best thing to do to make the fix. I don't know when I will figure this out or when it will appear in etch/proposed-updates. Again, sorry for the problem. carlo On Fri, 20 Apr 2007, M. Rovezzi wrote:
Quoting Bruce Ravel
: I don't quite understand this. Would a screenshot or a clar project file help?
Bruce,
I use Ifeffit and Horae packages under Debian built by Carlo so launching Artemis from the command line:
$artemis&
at each fit, during the minimization process, appear in the standard output, that is in the shell where I executed the program many lines (one for each path at each minimization step, if I well understand it) that have this form:
sum path #1 #2 #3 #4
Where #* are real numbers and sometimes appears "NAN" that, as Matt said, seems a division by zero or a log(-1). I'm trying to produce a project file that will show the case when the minimization fall in all "NAN" and the loop need to be killed to stop. I'll post it as soon as possible, but in the meantime is useful to understand these numbers that appear in the shell.
M.
-- Carlo U. Segre -- Professor of Physics Associate Dean for Special Projects, Graduate College Illinois Institute of Technology Voice: 312.567.3498 Fax: 312.567.3494 segre@iit.edu http://www.iit.edu/~segre segre@debian.org
participants (5)
-
Bruce Ravel
-
Carlo Segre
-
M. Rovezzi
-
Matt Newville
-
Mauro Rovezzi