Re: Displaying the gradient tolerance in the output

From: Chris Harrison (charris5_at_gmail.com)
Date: Thu Aug 23 2012 - 22:52:31 CDT

You have a few options:

1) You can output a forceDcd using the "forceDcdFilename" and
"forceDcdFrequency" parameters. You might then analyze the resultant
forces (note that you'll have the forces for each atom for each
outputted frame) looking for a point when they approach a minimum.
This is the more difficult but thorough of my suggestions.

2) You can load the namd log file in the VMD Plugin "NAMD Plot" and
output files for the energies that can be read into xmgrace or similar
tool. You can then calculate & plot the 2nd derivative of the
energies, particularly the potential energy, to acquire a
corresponding gradient.

Chris

On Tue, Aug 21, 2012 at 3:46 AM, Webber <wpliao_at_gmail.com> wrote:
> (Sorry I have a weird feeling that this email was not received by the
> mailing list (not showing up on the Mailing list webpage), thus sending it
> again, if you have received it I apologise for the double posting and please
> ignore this email, thank you)
>
> Hi all,
>
> After doing some more research on this topic I am starting to think that
> outputting the gradient tolerance may be a feature that has been taken out
> in recent versions (tutorial needs an update maybe?) and I guess that
> explains why no one can answer my question...
>
> If the energy gradient of iterations is not reported, may I ask how does
> someone determine whether a minimization or MD calculation has reached a
> level of convergence? I realise this may be a really silly question and MD
> calculations may not use energy gradient as the measure of convergence... If
> that is the case, what does a person write on a report or paper regarding
> the criteria of stopping the simulations? Right now, all I can think of is
> that "X was minimized with NAMD until the simulation reached convergence"
> which I don't think is very nice to read....
>
> Thanks for reading and all helps appreciated.
>
> Regards,
>
> Webber
>
>
> On 20 August 2012 10:02, Webber <wpliao_at_gmail.com> wrote:
>>
>> Hi all,
>>
>> I have been following the NAMD tutorial (ver: Feb 2012) to learn the
>> basics of
>> NAMD. When I was following the section on the output in the first chapter
>> (Basics of NAMD->Output: Water Sphere Log File, which can also be found at
>>
>> http://www.ks.uiuc.edu/Training/Tutorials/namd/namd-tutorial-unix-html/node10.html),
>> I found a discrepancy between the example log file, the portion of a log
>> file shown
>> on the tutorial page, and the log files I actually got. Both the example
>> log file and the
>> figure (fig 6) in the tutorial show lines describing the "initial step"
>> and the "gradient
>> tolerance", but these values are missing from my output. As the version of
>> the
>> tutorial is quite recent (Feb 2012), I don't think it is due to the
>> version of NAMD I use,
>> which is 2.9. (I also happen to have an older version (2.8) running on a
>> PPC, but
>> the output still doesn't have the values.)
>>
>> Though I think I follow the tutorial and the configuration file to minute
>> detail, I
>> wonder if I am missing some important settings, such as some options or
>> flags
>> I should be eliciting NAMD with, or some default settings for NAMD.
>>
>> Could someone please let me know how I can activate the display of the
>> gradient
>> tolerance? Or maybe shed some light on where best to look?
>>
>> Much appreciated.
>>
>> Regards,
>>
>> Webber
>
>
>

-- 
Chris Harrison, Ph.D.
NIH Center for Macromolecular Modeling and Bioinformatics
Theoretical and Computational Biophysics Group
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave., Urbana, IL 61801
http://www.ks.uiuc.edu/Research/namd       Voice: 773-570-6078
http://www.ks.uiuc.edu/~char                          Fax: 217-244-6078

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:57 CST