MathGroup Archive 1997

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

Search the Archive

Re: using indexed variables in MultiplierMethod[ ], etc.

  • To: mathgroup at smc.vnet.net
  • Subject: [mg7724] Re: using indexed variables in MultiplierMethod[ ], etc.
  • From: daiyanh at mindspring.com (Daitaro Hagihara)
  • Date: Wed, 2 Jul 1997 14:21:47 -0400 (EDT)
  • Organization: MindSpring Enterprises
  • Sender: owner-wri-mathgroup at wolfram.com

In article <5p75n1$8if at smc.vnet.net>, I wrote:

> SymbolicSum.  In the def. of SymbolicSum, the arguments are constrained
> by the Symbol heads.  So all I have to do to circumvent this is to fool
> the function as if symbolic arguments are being supplied.  If your
> function is doing the same thing as SymbolicSum, then this might help.
> 
> SymbolicSum[K,{K,n}] --> (n*(1 + n))/2
> SymbolicSum[K[1],{K[1],n}] --> SymbolicSum[K[1], {K[1], n}]
> SymbolicSum[K[1],{K[1],n}]/.K[1]->Symbol[K[1]] --> (n*(1 + n))/2
> 
> This may come in handy too:
> Thread[Array[u,2]->Array[Composition[Symbol,u],2]]
> --> {u[1] -> Symbol[u[1]], u[2] -> Symbol[u[2]]}
> 
> Ask the package author before attempting to modify the source code as
> there may be a reason why certain argument constraints are imposed.

Gee, the top portion of my previous post was gone!  It's because of my 
server, I believe.  Anyway, what I wrote was a comment, which is what I 
said in the blown opening statement of the previous post.  The problem 
with your MultiplierMethod, as it turned out, was the Compile function 
used in FindMinimum.  This Compile function simply doesn't accept 
non-Symbolic arguments.  Period.  You can turn off the auto compilation by 
using Compiled->False option in FindMinimum for using indexed variables, 
but this may not be desirable for a long and complicated cost function.  
The following will do the trick for you:

m:MultiplierMethod[_,_List,_List,x:{___,_?(Head[#]=!=Symbol&),___},_List,___]:=
(Hold[m]/.Thread[x->#])[[1]]/.Thread[#->x]&[If[Head[#]=!=Symbol,Unique[],#]&/@x];

Oh, if you define the above after loading the package, you must reverse 
the order of DownValues[MultiplierMethod]...a drag, but we must live with 
it.

--daiyan


  • Prev by Date: Re: nested * and ** (rules for commutative quantities)
  • Next by Date: Re: What is wrong?
  • Previous by thread: Re: Re: nested * and ** (rules for commutative quantities)
  • Next by thread: Package for Blockmatrices