| [eigen] anybody knows Japanese? |
[ Thread Index |
Date Index
| More lists.tuxfamily.org/eigen Archives
]
- To: eigen <eigen@xxxxxxxxxxxxxxxxxxx>
- Subject: [eigen] anybody knows Japanese?
- From: Benoit Jacob <jacob.benoit.1@xxxxxxxxx>
- Date: Thu, 30 Jul 2009 10:30:21 +0200
- Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:date:message-id:subject :from:to:content-type:content-transfer-encoding; bh=jh4WQXrx0yO3iUevhusDyr1GK8a+qnFL1I1nc+EEEO0=; b=U5g3XQlLpj3HTTyTb6aHuPkejmpGCRmyDe7nQnJc7/ZK81BJal4qdM7UG6/yxiwwwJ NUIg0GrHScNr9il0UAnTzPSQYLsn8UxQirCEp8qY29NIvmnta4BAmCCpgn0v6b0u6MTC hbVW5ZfYtmEyhQz0hOhdwrZlu0ml9CZK4RReg=
- Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type :content-transfer-encoding; b=Gdeu/Zhk4H4Jd9h1ISPyT792TZkoGApoZUsSQpMMfR4dp5H2FFIV/cmtewlb4cf4oF mgFWj4fmXihiDqPC/yVjHezZsT57yLu6yyiZEPv3VPYpz6vdKj/dQoMvVcFyvKu5eGrl L0DrJfXRxYcWaqAGF3ms1q7KpKf1JW8RJSekg=
Hi,
I'm finally redoing the SVD. As previously discussed, besides the
usual quick-and-dirty Golub-Kahan bidiagonalization, we also need a
pivoting variant. A quick google search suggests that it's well-known
that such a pivoting variant is needed in order to get numerical
accuracy comparable to Jacobi methods.
The problem is that the only paper that I found giving a pivoting
bidiagonalization algorithm is in Japanese:
http://ci.nii.ac.jp/lognavi?name=nels&lang=en&type=pdf&id=ART0002994867
Can anybody who knows japanese have a look at the second page and tell
roughly what the text in Step1--Step7 says?
Of course everybody can think of a simple and robust pivoting
strategy: just look for the biggest remaining row/column and swap it
with the current one. The problem with that method is that it's
expensive. Since these guys are making a paper about their algorithm,
I suppose that they had an idea to improve speed?
Cheers,
Benoit