[Date Index] [Thread Index] [Author Index]

Re: FindMinimum

• To: mathgroup at smc.vnet.net
• Subject: [mg51558] Re: [mg51547] FindMinimum
• From: DrBob <drbob at bigfoot.com>
• Date: Sat, 23 Oct 2004 00:21:39 -0400 (EDT)
• References: <200410220221.WAA07282@smc.vnet.net>
• Reply-to: drbob at bigfoot.com
• Sender: owner-wri-mathgroup at wolfram.com

If search routine moves along a gradient some discrete distance, it has no simple or fool-proof way of knowing whether it does, or does not, cross over a wall. It looks only at a finite number of points, so it can't really know the shape of the landscape.

Bobby

On Thu, 21 Oct 2004 22:21:33 -0400 (EDT), Steve Gray <stevebg at adelphia.net> wrote:

> 	FindMinimum is described as finding strictly LOCAL minima of the finction. To me this means
> that it will only go downward from the starting point. Therefore it should never cross "walls" or
> regions of higher value, even if the specified search limits includes such regions. That is what I
> wanted in a search of a highly nonlinear function of 2 variables. But I have some evidence that it
> did not strictly search downward, but rather found the lowest value in the search range by crossing
> over a region of higher value.
> 	Can anyone describe more thoroughly or accurately exactly what it does? Thank you.
>
> Steve Gray
>
>
>
>



-- 
DrBob at bigfoot.com
www.eclecticdreams.net

• References:
  • FindMinimum
    • From: Steve Gray <stevebg@adelphia.net>