From sources-request at octave dot org Fri May 28 22:04:21 2004
Subject: Re: Optimisation
From: Jskud at Jskud dot com
To: zenith at chemeng dot ntnu dot no
CC: sources at octave dot org
Date: Fri, 28 May 2004 20:03:44 -0700

Hi, Federico.

> I was wondering whether there was anybody working on optimisation
> algorithms.  lpsolve() answers still "not implemented yet" as of
> 2.1.57. Are they under active development?

I'm working on releasing an Octave version of Yinyu Ye's "solnp"
nonlinear solver under GPL.  solnp works well for (linear, quadratic,
and) non-linear programming problems.

It is based on his Stanford Ph. D thesis.  I've secured his permission
to release it under GPL, and am rewriting it for Octave (done),
documenting what's there (1/3 done) , adding a verification package
(done), and commenting the existing code (1/2 done).  There are no
substantive changes to the algorithm, so you can use his existing code
if you need a solver.

FYI, I implemented a complete method of multipliers for Octave (based on
what I did for Mathematica), but Yinyu Ye's solnp worked so much better
(easier to use, and faster in solving) that I've decided not to release
my "multmethod" code.

In addition, there is code available on source-forge for optimization.

So, yes, folks are working on optimization algorithms for Octave, and
working code for Octave is available on the web; it's just not (yet)
integrated with the Octave distribution.

/Jskud