[gmx-developers] Accurate energy minization is not reliable

[Erik Lindahl]

Hi,

Not really a short-cut - it's a general recommendation for all CG and
L-BFGS algorithms (i.e., not specific to GROMACS) to run the outer
iteration after a few steps instead of the line minimizer, since that is
much more efficient at finding a good local minimum.

If you have a concrete problem where it goes wrong put it on redmine and we
can try to have a look.

Cheers,

Erik

On Tue, Sep 11, 2018 at 11:02 AM David van der Spoel <spoel at xray.bmc.uu.se>
wrote:

> Hi,
>
> I am finding that energy minimization not always succeeds, even for very
> simple molecules such as 2-octanol or pentoxypentane. This leads to
> problems when you want to do a normal mode analysis based on that
> structure.
>
> The code for conjugate gradients has comments (from Erik) that suggest
> there are some short-cuts being taken (see example below). I do not want
> to modify this code, but it might be good to have a minimizer that
> *does* go all the way to zero force (steepest descents doesn't do it
> either).
>
> Suggestions?
>
>
>           * In theory, we should minimize the function along this
> direction.
>           * That is quite possible, but it turns out to take 5-10
> function evaluations
>           * for each line. However, we dont really need to find the
> exact minimum -
>           * it is much better to start a new CG step in a modified
> direction as soon
>           * as we are close to it. This will save a lot of energy
> evaluations.
>           *
>
>
> --
> David van der Spoel, Ph.D., Professor of Biology
> Head of Department, Cell & Molecular Biology, Uppsala University.
> Box 596, SE-75124 Uppsala, Sweden. Phone: +46184714205.
> http://www.icm.uu.se
> --
> Gromacs Developers mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-developers_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-developers
> or send a mail to gmx-developers-request at gromacs.org.
>


-- 
Erik Lindahl <erik.lindahl at dbb.su.se>
Professor of Biophysics, Dept. Biochemistry & Biophysics, Stockholm
University
Science for Life Laboratory, Box 1031, 17121 Solna, Sweden
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20180911/d7b32114/attachment.html>