FindMinimum
- To: mathgroup at smc.vnet.net
- Subject: [mg41235] FindMinimum
- From: Gabriel Molina-Terriza <gamote at ap.univie.ac.at>
- Date: Fri, 9 May 2003 03:20:02 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
Hello,
I am a newby in this group, so I hope to learn a lot about Mathematica and
help if possible.
Now, my problem. I am using FindMinimum to minimize a Likelihood function.
The function to be minimized has 81 variables and it is a sum of simpler
functions, which only contains polynomials or divisions of polynomials of
the variables. In any case, I don't know why, Mathematica is not able to
find the Gradient of this simpler functions. Another problem is that the
sum is a quite large one, it may contain up to 6500 elements (each element
being one of the simple functions). Do you think that Mathematica is the
right tool to make such a calculation? Is the sum too large for
Mathematica? Should I try harder to find the Gradient of the function?
Finally a more generic question, how can I take the output of FindMinimum
and use it for other variables?, For example the ouput is usually of the
form t1 -> 1.00, how can I make t1=1.00 or x=1.00, without typing it directly?
I attach part of the program I made, I save you the initialitation of the
variables:
...
FindMinimum[Re[L[t1,...,t81]][[1]],{t1,0,.1},...,{t81,0,.1},Evaluate[PrecisionGoal->7]]
Where, F is a trangular matrix with the variables, v is a list of the 6500
vectors, w is a list with the norms of The Fv vectors. Band Num is a list
of numbers, vartrac is the sum of the variables squared, and I think that's
all. If you find any better solution to all this lines, just tell me.
Thanks a lot!!!
Gabriel