Re: Optimization of simulation program result w.r.t. program parameters

[Francesco Potorti`]

>I need to optimize the final output (a real number) of a long program 
>w.r.t. three parameters that determine how the program works. I have 
>prepared the smallest toy example I could think of below. My purpose is 
>to compare the program's prediction of the final number of individuals 
>in a cohort (as a result of the parameters that determine egg 
>development and egg mortality in the early life) with the same number 
>obtained from another sampling-model process, to make these two numbers 
>as close as possible.
>I have found information to optimize a function (a mathematical 
>expression), for example with samin and bfgsmin, but I am not sure if 
>this is the right strategy to optimize the output of a long program. 

Yes, I think it is.  Even if I have used leasqr, never samin or bfgsmin,
I think that you are on the right track.

>my program, there are a few places where random numbers are called and 
>used to create variety among the individuals.

This will make convergence possible only if you are careful.  You cannot
ask for arbitrarily high accuracy.  You should try to evaluate what is
the dispersion of results due to the randomisation process, and ask for
an accuracy not smaller than that (in fact, reasonably higher than that).

-- 
Francesco Potortì (ricercatore)        Voice: +39 050 315 3058 (op.2111)
ISTI - Area della ricerca CNR          Fax:   +39 050 315 2040
via G. Moruzzi 1, I-56124 Pisa         Email: address@hidden
(entrance 20, 1st floor, room C71)     Web:   http://fly.isti.cnr.it/