From: Allen Abrahamson
Subject: Re: accuracy of computed matrix inverse
Date: Sun, 05 Nov 2000 03:40:32 GMT
Newsgroups: sci.math.num-analysis
Summary: [missing]
In article <8u2goa$ce1$1@nnrp1.deja.com>,
mittelmann@asu.edu wrote:
> You can get an exact inverse for nearly all matrices with iterative
> refinement, but that does require partial double precision.
Regarding the issue of iterative refinement, and its efficacy with
respect to precision of the original inversion: Say that we use the
iterative formulation (V_i are a series of numerical inverses, M is the
exact original matrix):
V_{i+1} = V_i + V_i (I - M V_i ).
The case is not doing a refinement of a linear solution vector, but only
of the inverse itself.
Question: If this iteration is to be carried out in precision d (.ge.
0) digits better than that used to obtain V_0, then, in general, what
may be said, a priori, about the ultimate "improvement", as measured by
properties of (I - M V_{i+1})?
Allen
------------------------------------------------------------------------
> In article ,
> prussing@aae.uiuc.edu (John Prussing) wrote:
> > One way to test the accuracy of a computed matrix inverse is obviously
> > to multiply the computed inverse by the original matrix and see how
> > close the product is to the identity matrix.
> >
> > 1) Is this the best way?
> >
> > 2) If some additional property exists, say the determinant is equal to
> > 1, can this be used to enhance the accuracy of the computed inverse?
> > --
> > =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> > John E. Prussing
> > Dept. of Aeronautical & Astronautical Engineering
> > University of Illinois at Urbana-Champaign
> > http://www.uiuc.edu/~prussing
> > =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
> >
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
>
Sent via Deja.com http://www.deja.com/
Before you buy.