MathGroup Archive 2008

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Do I have to worry about "Left" division for matrices?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg94658] Re: Do I have to worry about "Left" division for matrices?
  • From: dh <dh at metrohm.com>
  • Date: Fri, 19 Dec 2008 07:23:44 -0500 (EST)
  • References: <200812041217.HAA27795@smc.vnet.net> <ghdmo8$fjj$1@smc.vnet.net> <ghlmev$k5b$1@smc.vnet.net>


Hi,

I just saw your reply. Please note that using Inverse[A] to solve 

equations is numerically unstable. Use LinearSolve. Even  "A\vec" in 

your other language is only syntax. The real procedure does not directly 

use the matrix inverse.

The simplest procedure to solve equations is Gauss Elimination.

hope this helps, Daniel



hayes.tyler at gmail.com wrote:

> Thanks for all of the help and clarification. I like Daniel's

> solutionn best as it seems (to me at least) to be the most transparent

> method. Nonetheless, I may do a quick Timing[] test on Inverse[A].b

> versus LinearSolve[A,b] to see which performs best.

> 

> Cheers,

> 

> t.

> 

> 

> 

> 

> On Dec 6, 6:13 am, Daniel Lichtblau <d... at wolfram.com> wrote:

>> Tyler wrote:

>>> Hello All:

>>> I am currently in the process of rewriting some old code into

>>> Mathematica. One of the things I've come across is the matrix

>>> operation

>>> A \ b

>>> Which is defined as "left division." In other words (if I have this

>>> right):

>>> A \ b = inv(A)*b

>>> and "right division" is

>>> b / A = b*inv(A)

>>> I guess my question is, does mathematica make the distinction between

>>> left and right division, or is this more or less related to the

>>> algorithm used by the solver (I am using numerical values by the way,

>>> and not a symbolic solution)?

>>> Cheers,

>>> t.

>> The (preferred) Mathematica equivalents would be LinearSolve[A,b] and

>> LinearSolve[Transpose[A],b] respectively.

>>

>> Mathematica does not support an infix form of this sort of division, so

>> the immediate question is moot.

>>

>> Daniel Lichtblau

>> Wolfram Research

> 

> 




  • Prev by Date: Re: Definition of new distribution
  • Next by Date: Re: Elliptic Modular Function
  • Previous by thread: Re: Do I have to worry about "Left" division for matrices?
  • Next by thread: Re: Do I have to worry about "Left" division for matrices?