VMD-L Mailing List
From: Mayne, Christopher G (cmayne2_at_illinois.edu)
Date: Wed Apr 16 2014 - 09:44:19 CDT
My apologies for taking so long to respond to this email...proposal time around here.
It sounds like the SA algorithm either didn't find very good minima or very narrow minima, which the DH algorithm couldn't improve upon. One of things we're working on is speeding up some of the underlying calculations to enable much higher sampling in reasonable wall clock times, primarily for the benefit of the SA algorithm. This was the situation we experienced with the charge optimization routine very early ffTK's development. Our current practice is to run DH directly from the parameters extracted from the QM calculation (presumably close to a reasonable minimum), which has given us acceptable parameters thus far.
To judge the final objective value you can look into the ffTK LOG file written during the optimization. It contains contributions from each bond & angle. We are working on a modificatin of the COLP utility to plot this data in much the same manner as is done for charge optimization.
On Apr 7, 2014, at 2:42 PM, denilson_at_dqi.ufla.br wrote:
> Dear all,
> During the optimization of bond and angle parameters with FFTK in VMD 1.9.2a35 (binary files), which was installed on an Ubuntu 12.04.4 (64 bits) machine, initially I used the simulated annealing mode with T, Tsteps and Iter parameters set to 25, 100 and 100, respectively. This afforded bond and angle parameters with final objective value equal to 129. Then I used the results of this first calculation in the second calculation, which was carried out using the downhill mode with the Iter parameter set to 2000. The final objective value increased to 167 instead of decreasing to a value smaller than 129. Is it a normal behaviour ? By the way, how can I proceed to judge if the final objective value is acceptable ?
> Thanks in advance for any help.
> Denilson F. Oliveira
> Laboratório de Produtos Naturais
> Departamento de Química
> Universidade Federal de Lavras
> Caixa Postal 3037
> Lavras (A CIDADE DAS QUEIMADAS) - MG - BRASIL
> Fone: 55-35-3829-1623
> Fax: 55-35-3829-1271
> e-mail: denilson_at_dqi.ufla.br