Re: 3-D fit for extrapolation
- From: John D'Errico <woodchips@xxxxxxxxxxxxxxxx>
- Date: Tue, 30 Aug 2005 10:42:44 GMT
In article <1125393052.267819.243710@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
"Kimbusabi" <Gopal.K.Varma@xxxxxxxxx> wrote:
> I'm not entirely sure how to evaluate the the fit over the entire
> domain in the first method you described and when I tried using
> 'Inpaint_nons' downloaded from Matlab central it seemed to crash my
> computer. I would very much appreciate any further guidence if
> possible. A colleague suggested using something involving the simplex
> algorithm but in 2-D, any thoughts?
I don't know how large is the domain that you are
extrapolating over. (How much of that 512x270 grid
is given?) Inpaint_nans must solve a linear system
of equations in as many unknowns as you have nans.
So if there are 10000 total nans, then this is the
size of the system it must solve. Did it crash your
computer? Very likely not. Had there been a problem
then matlab would have given an out of memory error.
Instead, what happened is matlab was busy working.
It would have returned when it was done. In fact,
a problem with 10K unknowns is not at all uncommon
to solve, taking only a matter of seconds. In your
case, 512*270 = 138240. If 50% or those nodes are
known, then your system has 70k unknowns to solve
for. This is a moderately large problem, depending
on how much ram you have. Your computer may have
been going into virtual memory, which will be much
slower.
As for evaluating a polynomial fit over the entire
domain...
[x,y] = meshgrid(1:512,1:270);
Evaluate the polynomial for each (x,y) pair.
I have no idea what you friend was thinking about
with the simplex algorithm, especially since there
are at least two wildly different algorithms that
are often referred to as "the simplex algorithm"
that I can think of, and neither algorithm applies
at all directly to this problem.
John
--
The best material model of a cat is another, or
preferably the same, cat.
A. Rosenblueth, Philosophy of Science, 1945
.
- References:
- 3-D fit for extrapolation
- From: Kimbusabi
- Re: 3-D fit for extrapolation
- From: John D'Errico
- Re: 3-D fit for extrapolation
- From: Kimbusabi
- 3-D fit for extrapolation
- Prev by Date: Re: How to read matlab data into text file ?
- Next by Date: Re: Random Process Problem
- Previous by thread: Re: 3-D fit for extrapolation
- Next by thread: Re: spread spectrum system in jamming environments
- Index(es):
Relevant Pages
|