| Re: Any 1D functional minimization routines out there? [message #3860] |
Fri, 07 April 1995 00:00 |
volker
Messages: 3
Registered: April 1995
|
Junior Member |
|
|
Chris Daft (daft@debussy.crd.ge.com) wrote:
: Hi,
Hello
: I need a couple of functions corresponding to MNBRAK (golden section
: search for brackets for a function's minimum) and BRENT (1D
: non-derivative function minimization) in Numerical Recipes. I can't
: find these in the IDL distribution I use (3.6.1). Has anybody typed
: them in as IDL functions?
Why don't you try 'NR_POWELL'?
It's a function, which does the same than the powell-function in the
Numerical Recipies.
See IDL Mathematics Guide
ciao
volker
|
|
|
|
| Re: Any 1D functional minimization routines out there? [message #3867 is a reply to message #3860] |
Thu, 06 April 1995 00:00  |
geomagic
Messages: 22
Registered: November 1993
|
Junior Member |
|
|
In article <DAFT.95Apr4133459@debussy.crd.ge.com> daft@debussy.crd.ge.com (Chris Daft) writes:
dan> You might try FX_ROOT. It uses Muellers method.
dan> Dan (I speak for no organization) O'Connell
> Thanks for your post. I checked out Muller's method in the IDL docs
> and I understand what it does. I wonder if you, or someone else more
> versed in numerical analysis than I, could comment on this:
>
> Obviously, I can minimize my function by forming its derivative and
> root finding. But my function is not analytic and I am not keen on
> doing numerical differencing (because this is ill-conditioned, I
> imagine). Is there a deeper link between root finding and
> minimization which I am missing?
Muller's method does not require the evaluation of the function's
derivative and is global in the sense that the user need not supply
an initial approximation (in the IDL version you need to provide
3 non-equal numbers as the "initial guess" to get the routine started).
> If not, I guess I'll translate some C routines into IDL (I can't
> CALL_EXTERNAL, for another unrelated reason). But this will be slooow
> (loops) unless I spend a long time on it.
Try FX_ROOT first, it might be fine and it's easy to do. A more
general Muller's method alogorithm can be found in, Conte and de Boor's
elementary numerical analysis textbook. It does root deflation, so
it can return all possible roots.
Dan (I speak for no organization) O'Connell
geomagic@seismo.usbr.gov
Seismotectonics Group, U.S. Bureau of Reclamation
Denver Federal Center, P.O. Box 25007 D-8330, Denver, CO 80225
"We do custom earthquakes (for food)"
or
"Just more roadkill on the information superhighway"
/\
/ \
/ \ /\ /\
/\ / \ / \ / \ /\ /\/\ /\/\
___/ \ /\/\/\/ \ / \ /\ / \ / \/ \/ \ /\_______
\/ \ / \ / \/ \/ \/
\/ \/
|
|
|
|
|
|
| Re: Any 1D functional minimization routines out there? [message #3875 is a reply to message #3869] |
Tue, 04 April 1995 00:00 |
daft
Messages: 5
Registered: February 1991
|
Junior Member |
|
|
dan> You might try FX_ROOT. It uses Muellers method.
dan> Dan (I speak for no organization) O'Connell
Hi Dan,
Thanks for your post. I checked out Muller's method in the IDL docs
and I understand what it does. I wonder if you, or someone else more
versed in numerical analysis than I, could comment on this:
Obviously, I can minimize my function by forming its derivative and
root finding. But my function is not analytic and I am not keen on
doing numerical differencing (because this is ill-conditioned, I
imagine). Is there a deeper link between root finding and
minimization which I am missing?
If not, I guess I'll translate some C routines into IDL (I can't
CALL_EXTERNAL, for another unrelated reason). But this will be slooow
(loops) unless I spend a long time on it.
--
Chris M.W. Daft <daft@crd.ge.com>
KWC-1336, GE CR&D 518-387-6615
P.O. Box 8 518-387-7512 (fax)
Schenectady, N.Y. 12301-0008
|
|
|
|
| Re: Any 1D functional minimization routines out there? [message #3885 is a reply to message #3875] |
Sun, 02 April 1995 00:00 |
geomagic
Messages: 22
Registered: November 1993
|
Junior Member |
|
|
In article <DAFT.95Mar31084204@debussy.crd.ge.com> daft@debussy.crd.ge.com (Chris Daft) writes:
> I need a couple of functions corresponding to MNBRAK (golden section
> search for brackets for a function's minimum) and BRENT (1D
> non-derivative function minimization) in Numerical Recipes. I can't
> find these in the IDL distribution I use (3.6.1). Has anybody typed
> them in as IDL functions?
>
> If not, I guess I will use the C code and CALL_EXTERNAL (quite risky
> in my experience).
You might try FX_ROOT. It uses Muellers method.
Dan (I speak for no organization) O'Connell
geomagic@seismo.usbr.gov
Seismotectonics Group, U.S. Bureau of Reclamation
Denver Federal Center, P.O. Box 25007 D-8330, Denver, CO 80225
"We do custom earthquakes (for food)"
or
"Just more roadkill on the information superhighway"
/\
/ \
/ \ /\ /\
/\ / \ / \ / \ /\ /\/\ /\/\
___/ \ /\/\/\/ \ / \ /\ / \ / \/ \/ \ /\_______
\/ \ / \ / \/ \/ \/
\/ \/
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
