|Re: [eigen] Fast QR for 2x2, 3x3|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] Fast QR for 2x2, 3x3
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Mon, 23 Feb 2009 00:08:01 +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=1+OtDvLE7nnUVQhkos52SV+D/08jNqt0ax7hb5KSZ2Q=; b=riwXj1geHCv1YCllGZbp+b50Pf9zGs+0AMQKIW+JuGdKUyHMBHZcbiyYjlT45Cz6wY dCybIaItDrApy7olqp6VCfNNqOUFLjCeC1N/jfyZ5OTvQZi7bSK4vx7C5OnH5eYiPGY7 f6skSWrXtB6tQE175YFchAqDG2pj+tSGM70Oc=
- 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=lWfF2cyCQdCfe4taqfv0pzbjIND5qTJT8b4Q4YOPWUn5/Sz/YqPiahV6tezhICF2yN YNr+RTuAh5mrRdXBcPMo0MSmB8+HXbH/v9scbvs67lNIahxkrTVLZ23kOg73jpHVo0DR BSvVPcixUTyt0xN4EmF1mZLvJGXvgfrwx3zdw=
But indeed i was thinking (I could be wrong...) that for such small
sizes, nothing could beat just the plain direct Gram-Schmidt approach.
I'd be a little surprised if Givens rotations were faster.
2009/2/23 Benoit Jacob <jacob.benoit.1@xxxxxxxxx>:
> It would be very useful to have fixed-size specializations of QR for
> 2x2 and 3x3. You're very welcome to help us with that, and for that
> matter you can submit us the code in any format, it'll be easy to
> integrate in Eigen.
> I can't tell what's going to be faster, Givens or Householder or
> something else. All I can say for sure is that Givens rotations don't
> require computing cosines and sines!
> Here you can find a copy of Golub & van Loan: download.tuxfamily.org/eigen
> 2009/2/22 Andrea Arteaga <yo.eres@xxxxxxxxx>:
>> I'm Andrea Arteaga and I'm student of Computing Science and Engeneering at
>> I read this in the todo-page
>> implement efficient QR decomposition for 2x2 and 3x3 matrix with optional
>> computation of R (maybe using Gram-Schmitd instead of Householder
>> transformations ?)
>> I thought that Givens Rotations
>> (http://en.wikipedia.org/wiki/Givens_rotation) are perfect for that: the
>> computation of a givens-matrix requires only (execution of sin()) +
>> (execution of cos()) operations; the product between two 2x2 matrices
>> requires only 8 operations and maybe less than that.
>> If you need it, I can help a little bit, QR is one of my preferred topics!
>> [I'm sorry for my bad english, I live in Switzerland and I never studied
>> well english...]