| Re: Non-linear optimization [message #7430] |
Tue, 12 November 1996 00:00 |
Ken Kump
Messages: 8
Registered: February 1996
|
Junior Member |
|
|
Howard Onishi wrote:
> Does anyone know of a source for non-linear optimization codes (in IDL
> of course)? I am particularly interested in image restoration
> techniques such as maximum entropy deconvolution. The problem size is
> on the order of several hundred variables.
Howard,
Someone from the Naval Research Lab wrote a program called superfit.pro,
some years ago. It did a generalized non-linear least-squares fit to
a user-supplied function. This worked for me, but was quite slow. I
have found that for computationally intensive optimization, it is *much*
faster to use a call_external, especially if you are talking about
several hundred variables. I have used NL2SOL, a FORTRAN-based
procedure which is quite flexible and quite well documented. It can
be found at http://netlib.bell-labs.com. I have also used the
Powell-method (Numerical Recipies) in a similar fashion. Good luck.
------
Ken Kump
Department of Biomedical Engineering
Case Western Reserve University
Cleveland, Ohio 44106, USA
E-mail: ksk3@po.cwru.edu
|
|
|
|
| Re: Non-linear optimization [message #7434 is a reply to message #7430] |
Tue, 12 November 1996 00:00  |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
hto@kaiwan.com (Howard Onishi) writes:
>
> Does anyone know of a source for non-linear optimization codes (in IDL
> of course)? I am particularly interested in image restoration
> techniques such as maximum entropy deconvolution. The problem size is
> on the order of several hundred variables.
> Thanks,
> Howard Onishi
You might try an interesting new technique based on "pixons".
Dr. R. Puetter developed this technique primarily for deconvolution of
astronomical images, but you should be able to apply it to most
problems. His webpage:
http://cassfos01.ucsd.edu:8080/puetter.html
has references to the IDL source code as well as published articles
which you can download and read. The technique is iterative, a
portion of which is the standard maximum likelihood deconvolution.
You could probably strip this part out if you wanted to.
There are some restrictions for non-scientific use.
Best luck,
Craig
--
------------------------------------------------------------ ---------
Craig Markwardt UW-Madison 608 262 1164 | "To cogitate and
internet: craigm@astrog.physics.wisc.edu | to solve ..." -MathNet
------------------------------------------------------------ ---------
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
