Re: matrix inversion
- From: roberson@xxxxxxxxxxxxxxxxxx (Walter Roberson)
- Date: Thu, 24 Apr 2008 18:21:02 +0000 (UTC)
In article <fuqgr5$6pk$1@xxxxxxxxxxxxxxxxxx>,
Milad Ehtesham <m83ehtesh@xxxxxxxxx> wrote:
I have the following lower triangular matrix, which clearly
has determinant one and is full rank. but matlab gives zero
as its determinant and it cannot correctly inverted by
matlab. why this happens and how can i invert it correctly?
My hypothesis is precision limits and catestrophic cancellation.
Any 16 x 16 (or smaller) submatrix works; at 17 x 17 or larger
you start getting odd answers -- largely 0s, but some very
large and some very small.
rank() starts going wrong at 12 x 12
The matrix is badly conditioned according to rcond(). Even
just a 4 x 4 submatrix has an rcond of about 2E-6
--
"This quitting thing, it's a hard habit to break once you start."
-- Walter Matthau
.
- References:
- matrix inversion
- From: Milad Ehtesham
- matrix inversion
- Prev by Date: Re: ellipsoid
- Next by Date: how to draw the real time wave form
- Previous by thread: matrix inversion
- Next by thread: Re: matrix inversion
- Index(es):
Relevant Pages
|
