[Ifeffit] Random behaviour Artemis 0.7.001

Matt Newville newville at cars.uchicago.edu
Wed Mar 17 09:35:13 CST 2004


Mauro, Bruce, Peter,

> >>3. Sometimes the fit have no-sense values so I have to save the project,
> >>rerun Artemis, and with the same parameters I obtain good fit. It seems
> >>like after a certain number of fits the software starts to fail.
> > 
> > This one I *have* seen.  My current hypothesis is that I am using ifeffit's 
> > erase() command in a way that is flakey.  Unfortunately, I haven't been able 
> > to come up with a simple enough demonstration of the problem to merit 
> > complaining to Matt about it.  Indeed, I am not sure that erase() is really 
> > the problem.  In any case, I changed how erase() is used in a few places in 
> > the version I am currently working on and haven't seen this problem recently.  
> > So, there is hope ;-)  In the meantime, I don't have better advice than to 
> > save and restart.
> 
> I have also experienced the same problem when using "gifeffit", not artemis. 
> Sometimes a "mutant" of this bug appears with the fit parameters staying okay, 
> but the errors increasing with each fit attempt to enormous values. 
> Unfortunately I also didn't find a cause -> effect relation until now.

For what it's worth, I agree that erase() seems to cause problems on
some occassions, especially after multiple erase()'s and re-setting
of arrays.  I've also seen some odd problems using the slice()  and
interpolation functions in combination with the minimize() command.  
I have a hard time reproducing these problems myself, so they're
difficult to diagnose.

Sorry for the trouble!!

--Matt




More information about the Ifeffit mailing list