Re: minimum of a function into two variables



Andrea <acirulli@xxxxxxxxx> writes:

Hi,
I need some suggestion,
Let P_s=p!/{n choose p}

I need to calculate fixing P_s (for example P_s=2^-12) the minimum of
this function:
f(x)=p+ p \times \log{n}

There is no x on the right side of your equation.
Well, I suppose what you mean is you want to minimize
f(n,p) = p + p log(n) subject to the constraint p!/(n choose p) = c
where c is a positive constant. I'll assume 0 < c < 1.
Now we must be careful here, because normally p! and (n choose p)
are defined for nonnegative integers n and p (and so for all but
countably many c there would be no n and p satisfying the constraint),
but perhaps you mean a version of this that would be defined for real
n and p, presumably P_s(n,p) = Gamma(p+1)^2 Gamma(n-p+1)/ Gamma(n+1)
with n > 0 and n+1 > p > -1.

For fixed p > 0, P_s(n,p) ~ Gamma(p+1)^2/n^p -> 0 as n -> infty, while
P_s(n,0) = 1, so there is a curve p = g(n) for n sufficiently large
where 0 < g(n) < 1 and P_s(n,g(n)) = c. As n -> infinity, g(n) -> 0,
but rather slowly: it seems to me (using Maple) that
1/log(n) < g(n) < log(log(log(n)))/log(n)
for n sufficiently large. Now this means that
1 + 1/log(n) < f(n,g(n)) < (1 + 1/log(n)) log(log(log(n))).
I'm not convinced that f(n,g(n)) increases as n -> infinity, but if it
does then that happens only for very large n. For c = 0.8, the minimum
seems to occur around n = 300. For c = 1/2, it seems to be somewhere
between 500000 and 600000. For c = 2^(-12), I suspect the minimum would
be at a really enormous value of n, large enough that it will be extremely
difficult to find.
--
Robert Israel israel@xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Department of Mathematics http://www.math.ubc.ca/~israel
University of British Columbia Vancouver, BC, Canada
.



Relevant Pages

  • Re: MAPLE print content of myfile.m to maple standard output
    ... steineswz@hotmail.com (stijn) writes: ... > could not try your suggestion until now. ... make sure that maple hasn't read the m-file before you ... Joe ...
    (sci.math.symbolic)
  • Re: Firefox 1.0-2 crashes
    ... > Andrea Vettorello wrote: ... >> Have you tried what is suggested in the Debian Firefox documentation? ... I think you should follow Marc Wilson suggestion, ...
    (Debian-User)
  • Re: Calculus XOR Probability
    ... Thanks for the suggestion. ... size or infinity. ... have before it can be called an order relation on that set. ... Order isomorphisms are bijections, but so are other forms of ...
    (sci.math)
  • Re: nonlinear differential equation
    ... Maple says it's the inverse of an elliptic integral: ... hi Niel, ... I thank you for your suggestion, ... Posted Via Usenet.com Premium Usenet Newsgroup Services ...
    (sci.math)
  • [v3 PATCH 3/3] EDAC: CPC925 MC platform device setup
    ... This is the modified 3/3 patch that corrects CPC925 memory controller DTB node on Maple and setup related platform device for CPC925 EDAC driver. ... Michael's suggestion of being more discernible for the Maple platform only has been adopted. ...
    (Linux-Kernel)