|
From: | Arvid Rosén |
Subject: | Re: Extended precision |
Date: | Tue, 28 Feb 2006 07:48:40 +0100 |
User-agent: | Thunderbird 1.4 (Windows/20050908) |
Andy Adler wrote:
On Mon, 27 Feb 2006, Arvid Rosén wrote:Hi! I am frequently working with ill-conditioned state-space systems, and the matrix calculations often lead to problems. Are there any ways to get increasted or extended precision in Octave? 128-bit doubles or more would probably help alot in this case.I work with ill-conditioned systems for much of my research. and I would tend to be very suspicious of a statement like that. If the underlying problem is ill-conditioned then needing 128-bit doubles is probably avoiding the real issue, as an algorithm which is sensitive to rounding in this way has probably diverged long ago. I would look into regularization techniques to improve the conditioning of the algorithms. -- Andy Adler <address@hidden> 1(613)562-5800x6218
Thanks Andy!I'll bet you are right here. I was just hoping that someone would tell my about the new HYPER-MEGE-PRECISION-flag in octave-2.9 ;-) I guess I'll have to go and find myself a nice textbook (any recommendations?) on the topic and get started fixing the real problem.
/Arvid ------------------------------------------------------------- Octave is freely available under the terms of the GNU GPL. Octave's home on the web: http://www.octave.org How to fund new projects: http://www.octave.org/funding.html Subscription information: http://www.octave.org/archive.html -------------------------------------------------------------
[Prev in Thread] | Current Thread | [Next in Thread] |