MathGroup Archive 1999

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

Search the Archive

Problem -- perhaps memory -- with Table[ FindMinimum[] ]

  • To: mathgroup at smc.vnet.net
  • Subject: [mg17308] Problem -- perhaps memory -- with Table[ FindMinimum[] ]
  • From: "David Keith" <dkeith at sarif.com>
  • Date: Fri, 30 Apr 1999 23:22:33 -0400
  • Organization: Hevanet Communications
  • Sender: owner-wri-mathgroup at wolfram.com

I am trying to optimize a rather complicated function of 4 variables using
FindMinimum[]. (I won't try to show it here.) Since the function is not
differentiable, I am using the {x,{x1,x2}} formulation for starting values.

When run as FindMinimum[ f[a,b,c,d], {a,{a1,a2}}, . . . ]  I have no
problem. It yields a minimum in about 5 minutes. However, in order to look
for a best -- perhaps global -- minimum, I want to evaluate this within
Table[] like this:

    Table[
        FindMinimum[ f[a,b,c,d], {a,{a1,1.1 a1}}, . . . ] ,
        {a1, a1i, a1f, astep}, . . . . .
    ]

The idea, of course, is to construct a table of solutions based on an array
of starting values and examine them afterwards.

The problem is that, although the index values I'm using for Table[] require
only 2^4 = 16 evaluations, which should complete within 2 hours, after about
10 minutes the hard disk begins to work furiously, which I usually associate
with virtual memory.

Am I missing something here. The FindMinimum[] works fine by itself. When it
finishes the first one, Table[] needs so save a few hundred characters and
move on to the next.

Why the problem?

I'm running 3.0.1 on a PII/400 NT4/SP3 with 128MBytes. I have overwritten
mathkernel.exe with the out-of-memory patch.

Thanks.





  • Prev by Date: Y2K testing and time-expiring Mathematica passwords
  • Next by Date: Re: Complex Ratio Polynomial Fit
  • Previous by thread: Y2K testing and time-expiring Mathematica passwords
  • Next by thread: Re: Problem -- perhaps memory -- with Table[ FindMinimum[] ]