[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: LU - octave

From: David Doolin
Subject: Re: LU - octave
Date: Mon, 17 Jan 2000 12:21:15 -0500

In message <address@hidden>, "John W. E
aton" writes:
>On 17-Jan-2000, Igor Dukanovic <address@hidden> wrote:
>| I'm used to the following LU-factorization
>| PA = LU,
>| and so the octave's choice seems good for me.
>This is the way Octave does it because it is the way Matlab does it.
>No doubt there would be some complaints if Octave's choice were any

Several comments:  

Lapack was initially proposed by Demmel and Dongarra (lawn1 ---
lapack working note 1 available on netlib).

Demmel has co-authored pieces of Lapack software and the Lapack 

It turns out that a very beautiful proof of A = LU can be 
done inductively.  Demmel teaches it with complete pivoting:
PAP' = LU (Theorem 2.5, Applied Numerical Linear Algebra, Demmel,
1996).  Now an inductive proof may be beautiful in itself, but 
it turns out an algorithm for Gaussian Elimination is in
fact given by the induction step.  Clever, no?

Lapack extends Linpack, written in part by Dongarra and Cleve
Moler.   I am pretty sure Dongarra was a grad student with CM.
I am also pretty sure that CM == MathWorks == Matlab.

||P|| = 1  =>  P^T = P^{-1}, so there really isn't any 
inconsistency here: PA = LU  =>  A = P^TLU.

Matlab may or may not have distasteful inconsistencies, but 
this is not one of them.  


Dave D

>Octave is freely available under the terms of the GNU GPL.
>Octave's home on the web:
>How to fund new projects:
>Subscription information:

Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:
How to fund new projects:
Subscription information:

reply via email to

[Prev in Thread] Current Thread [Next in Thread]