Re: Matrix Inversion

From: Rich Townsend (rhdt_at_barREMOVEtol.udel.edu)
Date: 01/10/04

  • Next message: glen herrmannsfeldt: "Re: Newbie question about translating Fortran to C" 
    Date: Sat, 10 Jan 2004 01:38:58 -0500

    beliavsky@aol.com wrote:
    > porteiro-rs@bol.com.br (porteiro) wrote in message news:<1e208da0.0401091008.77497240@posting.google.com>...
    >
    >>Somebody can tell me how can I invert a matrix non-symetric and
    >>non-triangular using the ".i." operator?
    >>Thanks
    >
    >
    > As Dan Tex1 points out, Fortran 95 does not have an intrinsic ".i."
    > operator. The IMSL Fortran 90 MP LIBRARY that comes with Compaq Visual
    > Fortran, Profession Edition, does define ".i." to invert a matrix. So
    > you either write your own matrix inversion and overload it to ".i."
    > (which I recommend for portability) or buy CVF 6.6 Professional.
    >
    > Are you sure you need a matrix inversion procedure, not a linear
    > equation solver?

    If you write your own, a good place to start is Numerical Recipes in
    Fortran; online copies of the book can be accessed through the website
    at http://www.nr.com, although I myself find it more satisfying (and
    equitable) to actually own a copy of the book. The book provides a good
    introduction to the subject; however, its worth noting a couple of
    caveats before you take it as gospel truth. Firstly, while most of the
    algorithms are great for learning about numerical analysis, some ae not
    as up-to-date or accurate as they should be.

    Secondly, the copyright stipulations which appear in the book can appear
    rather threatening, and give the impression that anything based on their
    work is "derivative" and therefore illegal. Don't worry too much about
    this; the authors have made liberal use of other people's algorithms
    (the Brent root finder is a particularly-clear example; they haven't
    even bothered to change the variable names of the original ZEROIN
    function from Netlib), to such an extent that they are in a weak
    position to criticize others' use of the algorithms which they publish.
    Of course, this is not a recommendation to rip off their code verbatim,
    nor their explanatory text (which, IMHO, has a positive pedagogical
    value); but don't be afraid to implement you own routines based (and
    improved) on theirs. That's what their book is all about!

    cheers,

    Rich

  • Next message: glen herrmannsfeldt: "Re: Newbie question about translating Fortran to C"

    Relevant Pages

    • Re: branch and assignment statements translation to functional style
      ... the original algorithm i want to convert to functional ... from 'Numerical Recipes: The art of scientific ... Recipes has serious drawbacks as a source of numerical algorithms. ...
      (comp.lang.lisp)
    • Re: MSFT and random numbers
      ... algorithms in Numerical Recipes in C, hope the random number algorithm is ... See Howard Kaikow's web site. ... Randomize and Rnd in VB .NET ...
      (microsoft.public.dotnet.languages.vb)
    • Q: Good books for non-beginners
      ... Being the happy owner of a copy of the "Numerical recipes", ... am lost for algorithms. ... there are other books out there, which might supplement NR with more ... advanced algorithms, coverage in areas not dealt with by NR, or simply ...
      (sci.math.num-analysis)
    • Re: How to calculate IFFT based on FFT result?
      ... I just wonder if there is a simple way to calculate IFFT based on FFT ... "Numerical Recipes in C" explains some algorithms there. ...
      (comp.dsp)