|Re: [eigen] ordering of eigenvalues of EigenSolver|
[ Thread Index |
| More lists.tuxfamily.org/eigen Archives
- To: eigen@xxxxxxxxxxxxxxxxxxx
- Subject: Re: [eigen] ordering of eigenvalues of EigenSolver
- From: Susanne Suter <susanne.suter@xxxxxxxxx>
- Date: Tue, 22 Feb 2011 17:09:47 +0100
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:in-reply-to:references:date :message-id:subject:from:to:content-type:content-transfer-encoding; bh=OCIjH+muAdVYInLU8JGC+udbSTJSOaSDAqw7ms5ISLM=; b=KIfZk1Hg8kTK2BmXKI35HAdVtLBXR+F6yGQ9JATgzLbgmSJ9lsENc4jGFKRtQSrgIY SVZ4SA9oqfH8iHF8bHlY5caR1dyadHBTyan7ZxrOzQ5qplFIzxSY1Lg2BUAiF8rPQRbG IjB5lQ6g7tKIxumg+IObGtyb9j5w842Rbih4A=
- 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=ZOLaoeE8PXhORGn6s/5w+zEWt5x/Ur3jbcSIm8JmtNYj3yB2uqrdHHf9Yf/03bbD9a UlQTkdH2h+S0inaon3LiI0mvnVn7DFlrsOo312bOkgJIJBnvj+CsUkn2i+gYg29672LS Lr/syjPk4r1xyZjJ57wR8+p+ody56j+TK0uAc=
I'm sorry, I have to correct myself. Lapack does the same as eigen
regarding eigenvalue decomposition, so only MatLab does the sorting
for the x largest magnitude eigenvalues.
On Tue, Feb 22, 2011 at 8:57 AM, Susanne Suter <susanne.suter@xxxxxxxxx> wrote:
> thank you for taking the time to check that issue.
> I noticed now that I use instead of EIG/SSYEV the function EIGS in
> MatLab and the function SSYEVX in Lapack, which both return the x most
> significant eigenvalues. so I've learned by now that there are two (or
> more) types of eigenvalue decomposition methods provided in Matlab and
> Lapack while eigen supports one.
> thx again,
> On Tue, Feb 22, 2011 at 8:35 AM, Gael Guennebaud
> <gael.guennebaud@xxxxxxxxx> wrote:
>> On Tue, Feb 22, 2011 at 1:18 AM, Benoit Jacob <jacob.benoit.1@xxxxxxxxx> wrote:
>>>> Furthermore, other eigenvalue decomposition libraries seem to order
>>>> eigenvalues, too. E.g., MatLab (based on ARPACK)
>>> Ah, I didn't know about Matlab.
>> actually MatLab does as us and Lapack do, for instance:
>> $ eig( diag([1 -1 2 -3 4]))'
>> ans =
>> -3 -1 1 2 4
>> Same for Octave or Scilab.
>> I don't see any reason to change for a convention that nobody follow
>> while our current behavior seems to be quite standard.
>> On the other hand we could allow for sorting capability such that user
>> can easily sort rows or column as they like.