From: Tamas
Subject: multivariate optimization (quasi-Newton)
Date: Mon, 03 Dec 2007 01:28:24 +0000
Message-ID: <aaebb20e-c580-44bf-9b34-d91fec7f9d86@d4g2000prg.googlegroups.com>

Hi,

I am looking for multivariate optimization code in Lisp.  I found some
Nelder-Mead algorithms, but I need methods that use derivatives (BFGS,
etc).  Of course the matrix updating part is easy, but efficient line
search algorithms are full of special cases and hell to write.

If anyone has code or pointers, it would help a lot.

Thanks,

Tamas

From: Robert Dodier
Subject: Re: multivariate optimization (quasi-Newton)
Date: Mon, 03 Dec 2007 04:18:02 +0000
Message-ID: <827ee2f3-ccce-4205-9415-92b6522b9479@a39g2000pre.googlegroups.com>

Tamas wrote:

> I am looking for multivariate optimization code in Lisp.  I found some
> Nelder-Mead algorithms, but I need methods that use derivatives (BFGS,
> etc).  Of course the matrix updating part is easy, but efficient line
> search algorithms are full of special cases and hell to write.

There is a translation (from Fortran via f2cl) of the LBFGS
(limited-memory BFGS) algorithm in Maxima.
See share/lbfgs in the Maxima distribution.
?? lbfgs
at the Maxima input prompt shows documentation topics.
i've used LBFGS for various purposes and it always works OK for me.

HTH

Robert Dodier