[gmx-users] Steepest descent and max(|F|)

[Marco Ceruso]

Hi-
In steepest descent minimization, the step is taken straight down hill (-Fn) but they are several flavors for the step size:
fixed, variable, taken to depend on the magnituge of the gradient. The latter seems to be the case here.
 From what I understand in the code (minimize.c) at a given step in the minimization there is a stepsize, ustep, which is the full/maximum stepsize, that is constant and that is divided by max(|Fn|) this makes sure that the full/max stepsize is taken only by the coordinate that has the steepest gradient. So the gradient dependent step for coordinate n is hn*Fn with hn=ustep/max(|Fn|).

Marc
 
PS: it appears that contrary to what the manual says ustep is multiplied by 0.5 (code) rather than 0.2 (manual) when the new energy is higher.

-----Original Message-----
From: gmx-users-admin at gromacs.org [mailto:gmx-users-admin at gromacs.org]On
Behalf Of s8026264
Sent: Monday, September 01, 2003 2:11 AM
To: gmx-users at gromacs.org
Cc: mceruso at physbio.mssm.edu; lindahl at stanford.edu
Subject: RE: [gmx-users] Steepest descent and max(|F|)


Dear All

My purpose is page 33 of the manual;part 3.10.1 steepest descent.

On Sun, 31 Aug 2003 18:15:36 -0400, Marco Ceruso wrote
> Hi-
> When you are minimizing a system with multiple atoms you are 
> sometimes interested in knowing that the maximum force in your 
> system is below a certain threshold. I hope this answers your 
> question, since I am not really sure which part of the manual you 
> are referring to. Marc
> 
> -----Original Message-----
> From: gmx-users-admin at gromacs.org [mailto:gmx-users-admin at gromacs.org]On
> Behalf Of s8026264
> Sent: Sunday, August 31, 2003 7:35 AM
> To: gmx-users at gromacs.org
> Subject: [gmx-users] Steepest descent and max(|F|)
> 
> Dear All
> 
> In the manual you use max(|Fn|),Why don't you use |Fn| ?
> 
> The notation max(|Fn|) means the largest of the absolute values of 
> the force components.
> 
> The notation |Fn| means the absolute values of the force.
> 
> Good Luck
> 
> Mojtaba Alaei
> 
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
> 
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.


--

_______________________________________________
gmx-users mailing list
gmx-users at gromacs.org
http://www.gromacs.org/mailman/listinfo/gmx-users
Please don't post (un)subscribe requests to the list. Use the 
www interface or send it to gmx-users-request at gromacs.org.