Re: Geometric average: how to compute it: best approach ?
- From: Duncan Muirhead <dmuir@xxxxxxxxx>
- Date: Tue, 17 Jul 2007 17:48:09 +0100
On Tue, 17 Jul 2007 05:03:44 -0700, beeworks wrote:
On Jul 17, 7:16 am, Raymond Manzoni <raym...@xxxxxxx> wrote:<snip>
pamela fluente a écrit :
A more general recursion is derived as follows. Let g() be theI don't understand the last part. It seems to me that the number of
function that returns the geometric mean of its arguments. Then
g(x_1, x_2, ..., x_n) = (x_1 * x_2 * ... * x_n)^(1/n)
= (x_1 * x_2 * ... * x_m)^(1/n) * (x_(m+1) * x_(m+2) * ... * x_n)^(1/
n)
= g(x_1, x_2, ..., x_m)^(m/n) * g(x_(m+1), g_(m+2), ..., x_n)^((n-m)/
n)
We can apply the formula above recursively to find the g.m. while
multiplying only two numbers together at each step. Furthermore,
without loss of generality, we can assume that
x_1 <= x_2 <= ... <= x_n
Then
x_1 <= g(...) <= x_n,
where g(...) represents any of the intermediate g.m.'s. Thus the
formula, applied recursively, never has to deal with a result larger
than (x_n)^2 and the number of exponents n_e that must be found lies
between log_2(n) and 2*log_2(n), which is much less than n.
- MO
exponentiations will be of the same order as the number of data.
If E(n) is the number of exponentiations then if the recursion is
g(x_1, x_2, ..., x_n) =
g(x_1, x_2, ..., x_m)^(m/n) * g(x_(m+1), g_(m+2), ..., x_n)^((n-m)/n
then we get E(n) = E(m) + E(n-m) + 2
with E(1) = 0 we get E(2^n) = 2^(n+1)-2
Even in the ideal case where n is a power of 2 and the recursion is
g(x_1, x_2, ..., x_2n) =
sqrt( g(x_1, x_2, ..., x_n) * g(x_(n+1), g_(m+2), ..., x_2n))
we have E(2n) = 2*E(n)+1
and we get E(2^n) = 2^n-1
.
- Follow-Ups:
- Re: Geometric average: how to compute it: best approach ?
- From: pamela fluente
- Re: Geometric average: how to compute it: best approach ?
- References:
- Geometric average: how to compute it: best approach ?
- From: pamela fluente
- Re: Geometric average: how to compute it: best approach ?
- From: Raymond Manzoni
- Re: Geometric average: how to compute it: best approach ?
- From: beeworks
- Geometric average: how to compute it: best approach ?
- Prev by Date: Re: Ultimate debunking of Cantor's Theory
- Next by Date: Grand Successor Conjecture.By Aiya-Oba
- Previous by thread: Re: Geometric average: how to compute it: best approach ?
- Next by thread: Re: Geometric average: how to compute it: best approach ?
- Index(es):
Relevant Pages
|