P-dimen min of function, NR in C, powell() - anyone got it to run properly? - Thread - go-pugleaf

Thread View: sci.math.num-analysis

2 messages

[T] Back to Threads [V] Tree View [G] Group Articles

2 total messages Started by Jeff Racine Thu, 01 Oct 1992 19:11

P-dimen min of function, NR in C, powell() - anyone got it to run properly?

#706

Author: Jeff Racine
Date: Thu, 01 Oct 1992 19:11

23 lines
1402 bytes

Hi.

I am trying to minimize a multidimensional function of 5 arguments
without using derivatives. Numerical recipes in C has a routine called
powell(). I have it up and running, but one problem remains.

The routine works fine for 1 dimension, but for two or more
dimensions, it yields the same parameter values for each dimension. I
know this is not the case because simple grid search yields a much
lower value of the objective function with parameters being much
different.

Question: has anyone got the algorithm to work properly? I am using
the 1988 version of NR and associated code. If you have, would you let
me know - I would like to discuss this with you. To get the code
running on an ANSI compiler I had to make a few minor changes, and it
is possible that here is where the bugs crept in.

-- 
Jeff Racine        
jracine@weber.ucsd.edu
racine@nexus.yorku.ca
--Remember when sex used to be safe and parachuting was dangerous?

Re: P-dimen min of function, NR in C, powell() - anyone got it to run properly?

#714

Author: furio ercolessi
Date: Fri, 02 Oct 1992 04:35

17 lines
1392 bytes

In article <1afig6INNapa@network.ucsd.edu>, jracine@weber.ucsd.edu (Jeff Racine) writes:
|>I am trying to minimize a multidimensional function of 5 arguments
|>without using derivatives. Numerical recipes in C has a routine called
|>powell(). I have it up and running, but one problem remains.
|>  [... description of bug ...]

well the fortran version works (although it's slowwww on my
40-parameters problem).
of course i am not saying this to raise the silly language war,
but rather to suggest that perhaps comparing the two versions
may help in tracking the bug. In fortran it's 40 lines long.
furio
--
Furio Ercolessi
Materials Research Laboratory           |   Intl School for Advanced Studies
Univ. of Illinois at Urbana-Champaign   |   Trieste, Italy
furio@uiuc.edu                          |   furio@sissa.it

Thread Navigation

This is a paginated view of messages in the thread with full content displayed inline.

Messages are displayed in chronological order, with the original post highlighted in green.

Use pagination controls to navigate through all messages in large threads.

Back to All Threads