Re: [bug #11454] Constraint incorrectly supplied when using the "full_analysis.py" script -- April 08, 2008 - 13:31

Hi Ed,

These news concerning the minimization of model "m0" are reassuring. I agree with you that keeping this behaviour is better than trying to change it and potentially introducing bugs... However, I would agree if a small change would be made so the error message says the user that, in this special case of "m0", the error is normal.

Also, and more concerning, is the apparent break in the monte carlo simulations... Please see the second message in the bug report (11454) where I explain that the log file for any given simulation stops after encountering the "m0" problem even though the results file seems to contain the simulations results for every residue... This is quite strange and I'm afraid that the simulations might be affected... Do you think that the simulations for residues after the "m0" problem will be fine ? If so, why isn't there anything in the logs concerning residues following the first "m0" problem..?

Cheers


Séb






Edward d'Auvergne wrote:

Hi,

For model m0, this shouldn't be a problem.  This model has no
parameters, hence the dimension of A is 0x0 and for b is 0 (these are
the constraint matrices such that A.x <= b, where x is the parameter
vector).  Therefore it doesn't mater if minimisation for the
individual model-free model fails, because there is nothing to
minimise.  Code could be added to catch this model and skip it, but
I'd prefer that this behaviour stays.  Adding new code for model m0
could introduce bugs, whereas the current behaviour means nothing
happens for this model.  So this message will appear for the normal
optimisation of m0 and the Monte Carlo simulations, but is harmless.
Does this message appear for other model-free models?

Cheers,

Edward


On Tue, Apr 8, 2008 at 4:05 AM, Sébastien Morin
<NO-REPLY.INVALID-ADDRESS@xxxxxxx> wrote:
  

 Follow-up Comment #2, bug #11454 (project relax):

 Hi,

 I had a look at the log files for other runs and found out that the same
 error message appeared for minimization of model "m0" for every diffusion
 tensor. For example, here's what's found for the prolate diffusion tensor
 right after initialization (folder "init") :


 ============================
 Method of Multipliers
 ~~~~~~~~~~~~~~~~~~~~~
 The constraints have been incorrectly supplied.

 relax> results.write(run='m0', file='results', dir='prolate/round_1/m0',
 force=1, format='columnar', compress_type=1)
 Opening the file 'prolate/round_1/m0/results.bz2' for writing.
 ============================


 The error message appears for every round when at model "m0".


 For runs without model "m0", the error message never arises...




 Séb

    _______________________________________________________

 Reply to this item at:

  <http://gna.org/bugs/?11454>

 _______________________________________________
  Message sent via/by Gna!
  http://gna.org/