MathGroup Archive 2008

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

Search the Archive

Re: symbolic matrix inverse

  • To: mathgroup at smc.vnet.net
  • Subject: [mg94227] Re: symbolic matrix inverse
  • From: magma <maderri2 at gmail.com>
  • Date: Mon, 8 Dec 2008 06:23:36 -0500 (EST)
  • References: <ghb014$ogk$1@smc.vnet.net>

On Dec 5, 11:33 am, Steven Allmaras <steven.r.allma... at boeing.com>
wrote:
> I have a symbolic matrix {{a,b},{c,d}} whose elements are themselves matr=
ices.  The elements are noncommutative (e.g., a.b != b.a), and just for=
 simplicity assume each element is a n-by-n square matrix of reals.  How =
do I get Mathematica to evaluate Inverse[{{a,b},{c,d}}] in terms of Dot and=
 Inverse, rather than Times and Divide?  I'm hoping to get something that=
 looks like,
>
> [[1,1]] = Inverse[a] + Inverse[a].b.Inverse[d - c.Inverse[a].b].c.Inver=
se[a]
>
> Thanks,
> Steve

I do not think you can overload the existing Inverse function so that
it works with Dot or Composition or that in general that it works with
matrices in a non-commutative ring.
I believe the only solution is to define your own Inverse function
(better perhaps call it something else) along the outline that you
posted.
I suggest that you consider looking at Composition , InverseFunction
and Working with Operators sections in the documentation.

hth


  • Prev by Date: Re: Subscripts in version 7
  • Next by Date: Re: Clever Tricky Solutions
  • Previous by thread: RE: symbolic matrix inverse
  • Next by thread: Re: InverseLaplaceTransform different solution in Mathematica 6 and