MathGroup Archive 2008

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

Search the Archive

Re: Problems with FindMinimum

  • To: mathgroup at
  • Subject: [mg90724] Re: Problems with FindMinimum
  • From: Jean-Marc Gulliet <jeanmarc.gulliet at>
  • Date: Mon, 21 Jul 2008 04:29:40 -0400 (EDT)
  • Organization: The Open University, Milton Keynes, UK
  • References: <g5v5mf$ia9$>

teodorom at wrote:

> I'm having problems with the FindMinimum.
> I'm currently getting errors like this:
> FindMinimum::nrlnum: The function value {0.335521+43899.8 \
> [ImaginaryI],0.376043+43899.8 \[ImaginaryI],0.424669+43899.8 \
> [ImaginaryI],<<6>>,0.311208+46311.9 \[ImaginaryI],<<27>>} is not a
> list of real numbers with dimensions {37} at {x0,y0,x1,y1,hh0,hh1} =
> {0.00580647,-2.13969*10^20,-9.86301*10^-6,1.23875*10^20,-30574.3,19.6042}.
> I guess this happens because the function I'm trying to fit with
> experimantal data, gets imaginary values for some values of the
> parameters.
> How can I avoid this and get meaningful results ?

Hard to tell w/o more information about what your code actually is. It 
would have been very helpful for the group if you had posted an small 
self-contained fully-working example that reproduces the error and gives 
the gist of what you are doing.

Anyway, the following tip might help you. Sometimes, it is a good idea 
to write a model in a slightly different way to force Mathematica to use 
real numbers only. For instance, evaluate and compare the following 
expressions that all deal with the same function (cubic root of a 
square) but written in different fashions. (The graphics have been deleted.)

     f = x^(2/3);
     Plot[f, {x, -1, 1}]
     FindMinimum[f, {x, 1}]

     During evaluation of In[1]:= FindMinimum::nrnum:The function value
     -0.381571+0.660901 I is not a real number at {x} = {-0.666667}.

   Out[3]= {1., {x -> 1.}}

     f = (x^2)^(1/3);
     Plot[f, {x, -1, 1}]
     FindMinimum[f, {x, 1}, WorkingPrecision -> 20] // Chop

     During evaluation of In[4]:= FindMinimum::lstol:The line search
     decreased the step size to within tolerance specified by
     AccuracyGoal and PrecisionGoal but was unable to find a sufficient
     decrease in the function.  You may need more than 20.` digits of
     working precision to meet these tolerances.

   Out[6]= {0, {x -> 0}}

-- Jean-Marc

  • Prev by Date: Re: export eps
  • Next by Date: Re: On default numerical precision
  • Previous by thread: Re: Problems with FindMinimum
  • Next by thread: RE: Problems with FindMinimum