Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.


fd
Posts:
43
Registered:
7/24/09


Explanation to Hold?
Posted:
Nov 18, 2010 7:06 AM


Dear Group,
While doing some work in Mathematica I stumbled into something that has let me quite intrigued. Indeed, I'm having some difficulties to understand how hold works. I'm an unexperienced user, so any clarification on this would be greatly appreciated.
I'm using Mathematica to generate a report where I display points of minimum for two different function and a set of 4 parameters.
I start by defining the functions like this
fnA = Function[{x}, x^2 + x]
fnB = Function[{x}, x^4  x^2]
results = {fnA, fnB};
pointsOfInterest = {1, 2, 3, 4};
And then I generate the points of minimum using the following command
Thread[NArgMin[#, x] & /@ ((#  (pointsOfInterest))^2) & /@ Through[results[x]]] // Grid
All works fine, but then, when I use compiled functions
fnAC = Compile[{x}, fnA[x]];
fnBC = Compile[{x}, fnB[x]];
resultsC = {fnAC, fnBC}
and use the same command as before
Thread[NArgMin[#, x] & /@ ((#[x]  (pointsOfInterest))^2) & /@ resultsC]
I get the following message, along with the answer I had before.
CompiledFunction::cfsa: Argument x at position 1 should be a machine size real number.
Well, I thought that functions like NArgMin would first assign a real value to x and then evaluate the function, as Plot would do. When I hold the compiled function
NArgMin[Hold[(fnAC[x]  3)^2], x]
I do get the result without any warning message.
If I plot the compiled functions I don't get any message either.
Funnier still, when I use the mapping to plot everything at once
Thread[Plot[#, {x, 10, 10}] & /@ ((#[x]  (pointsOfInterest))^2) & /@ resultsC]
I do get the same message
CompiledFunction::cfsa: Argument x at position 1 should be a machine size real number
But in this case I can remedy by using Hold and ReleaseHold
Thread[Plot[ ReleaseHold[#], {x, 10, 10}] & /@ ((Hold[#[x]]  (pointsOfInterest))^2) & /@ resultsC]
But this same trick will not work with NArgMin, which is puzzling to me.
Well, any comments?
Best Regards



