[Dock-fans] Something about minimization

[Scott Brozell]

Hi,

On Tue, 25 Mar 2008, Äî Áõ wrote:

>   I noticed something about minimization in dock user manual. In "Score Optimization Parameters" , I found that simplex_max_iterations parameter and simplex_max_cycles parameter have the same descriptions that is maximum number of minimization cycles. I'm puzzled about it and expecting someone could interpret it.
> Yolanda Guo
> Northeast Normal University   

simplex_max_iterations
specifies the convergence criterion in number of iterations inside
one call of the simplex minimizer; the other convergence criterion
inside one call of the simplex minimizer is simplex_score_converge.

Because of convergence issues with the downhill simplex method,
additional calls of the simplex minimizer inside one step of
minimization are possible: these convergence criteria are
simplex_max_cycles for the maximum number of calls, aka cycles,
of the simplex minimizer and simplex_cycle_converge for the 
normalized vector distance that the simplex has moved.
http://dock.compbio.ucsf.edu/DOCK_6/dock6_manual.htm##EwingetalJCAMD2001
See page 416 and 
note that the citation numbers are off by one: i.e., [24] should be [23].

The manual could be clearer and more complete; it has been updated.

Note that all the above applies to other uses of the simplex minimizer
simplex_anchor_max_iterations, ...
simplex_grow_max_iterations, ...
simplex_final_max_iterations, ...

Scott

This article was not much help:
http://en.wikipedia.org/wiki/Nelder-Mead_method