| Re: [eigen] Levenberg-marquardt and dogleg minimizer |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Levenberg-marquardt and dogleg minimizer
- From: Gael Guennebaud <gael.guennebaud@xxxxxxxxx>
- Date: Thu, 26 Feb 2009 11:57:43 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:content-type :content-transfer-encoding; bh=4nQoZ1pyWDu61V/eNswXlonBJlQFF99VjtIIg9K4hjE=; b=oFqDq+bYCVge9vxao6nSG1nz89JtYQssnCccdv6W/r/PyJ2Xmzecuk/2Sxj3Cy3yde v/nvnM1goDKwp7faW/H38FMK0EYexEfYQ/8q+k9JX49yODBLW0SKSD50jYRJakCA5mb3 mudf5nxbOnzdHLtki2NDzTtaJYYPW/NArIV+k=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :content-type:content-transfer-encoding; b=Tpw2ZooShA0L000xfixFjwhpgRXhN4OAec8JpAURPENJD1G54vTaMcguib2S/Olu4o ztDqSssf2BI+Z/ovcTfMpRnLDsdJszr1auQVptOGuaQ/OVGI0TA5P1XS8nMT01BsO5TQ KZNjd/WJlDnc2Rr9lq7NnRjhTGzZ80iSG8czs=
On Wed, Feb 25, 2009 at 8:10 PM, Jos van den Oever <jvdoever@xxxxxxxxx> wrote:
> 2009/2/25 Gael Guennebaud <gael.guennebaud@xxxxxxxxx>:
>> as soon as I have little time I'll try to implement such an autodiff
>> Jacobian class using adolc to see what are the exact requirements
>> (according to my experience with adolc this should be easy, but we
>> never know).
>
> I would certainly be interested in such a class. This would be for a
> problem where we use a different solver which also requires a
> Jacobian. This solver can take advantage of the sparseness of the
> Jacobian. The sparseness should be set in advance. Also if it is known
> that a parameter has a linear influence, this can help the iterations.
> Would an adolc Jacobian as you envision it, know these properties?
adolc internally uses dense vector to store each derivatives, so it
cannot deal with large - sparse problem. What would be nice is to
implement our own autodiff tool based on eigen's dense and sparse
objects but that happens only if someone interested in doing that
shows up...