Part of the reason I even ever tried CG is that I occasionally have
problems with L-BFGS (i.e. it stops after zero steps and says it's
done) and I need to use another minimization method. But I guess rather
than complaining about CG I should submit a Bugzilla so you can
reproduce my problems with L-BFGS. :) I'll try to get my act together
and do this soon.<br>
<br>
Thanks,<br>
David<br>
<br><br><div><span class="gmail_quote">On 12/16/05, <b class="gmail_sendername">Erik Lindahl</b> <<a href="mailto:lindahl@sbc.su.se">lindahl@sbc.su.se</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi,<br><br>Mark is mostly right - the main reason is that we simply haven't<br>found it worth the effort yet :-)<br><br>The underlying problem is that minimization is really sensitive to<br>tiny differences in coordinates, and even conservative constraint
<br>algorithms usually leave constraint errors in the order of 1e-6. This<br>can lead to really strange artifacts since your system is no longer<br>strictly conservative, so if you want really good minimization the<br>best bet is to allow the molecule to be flexible and use
e.g. BFGS.<br><br>We're aware of it though, and will eventually add code to project out<br>constraint forces with CG too, but for the above reason it's not a<br>top priority!<br><br>Cheers,<br><br>Erik<br><br><br>On Dec 16, 2005, at 11:42 PM, Mark Abraham wrote:
<br><br>> David Mobley wrote:<br>>> Mark,<br>>> It may well be the case that implicit water constraints are the<br>>> problem, as the follow-up response seems to indicate (I haven't<br>>> tried it myself). I guess I'd suggest that this be stated
<br>>> explicitly in the manual (and perhaps even in the section on input<br>>> files which discusses options for minimizers, under the CG<br>>> section, there should be a note saying that CG won't work with
<br>>> rigid water...).<br>><br>> I filed a Bugzilla report for the same at http://<br>> <a href="http://bugzilla.gromacs.org/show_bug.cgi?id=37">bugzilla.gromacs.org/show_bug.cgi?id=37</a><br>><br>>> Also, perhaps this is naive, but CG works fine in AMBER without
<br>>> doing anything special. It might be worth checking there to see<br>>> how AMBER handles this. Perhaps it is "properly" handling the<br>>> water, and if so, GROMACS could do the same thing? Or perhaps it
<br>>> lets the water be flexible? I'm not sure, but it seems like the<br>>> issue must already be one people have dealt with in these other<br>>> packages.<br>><br>> It's probably just that constraints aren't implemented in the
<br>> gromacs CG algorithm. It's probably not worth implementing either,<br>> given that constraints can be used with steepest descent and L-BFGS.<br>><br>> Mark<br>> _______________________________________________
<br>> gmx-users mailing list<br>> <a href="mailto:gmx-users@gromacs.org">gmx-users@gromacs.org</a><br>> <a href="http://www.gromacs.org/mailman/listinfo/gmx-users">http://www.gromacs.org/mailman/listinfo/gmx-users
</a><br>> Please don't post (un)subscribe requests to the list. Use the www<br>> interface or send it to <a href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a>.<br><br>-----------------------------------------------------------
<br>Erik Lindahl <<a href="mailto:lindahl@sbc.su.se">lindahl@sbc.su.se</a>> Backup address:<br><<a href="mailto:erik.lindahl@gmail.com">erik.lindahl@gmail.com</a>><br>Assistant Professor, Computational Structural Biology
<br>Stockholm Bioinformatics Center, Stockholm University, SE 106 91<br>Stockholm<br>Phone: +46 8 5537 8564 Fax: +46 8 5537 8214<br><br><br><br><br>_______________________________________________<br>gmx-users mailing list
<br><a href="mailto:gmx-users@gromacs.org">gmx-users@gromacs.org</a><br><a href="http://www.gromacs.org/mailman/listinfo/gmx-users">http://www.gromacs.org/mailman/listinfo/gmx-users</a><br>Please don't post (un)subscribe requests to the list. Use the
<br>www interface or send it to <a href="mailto:gmx-users-request@gromacs.org">gmx-users-request@gromacs.org</a>.<br></blockquote></div><br>