MathGroup Archive 2005

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

Search the Archive

Re: [TS 28968] Mathematica SingularValueDecomposition failure

  • To: mathgroup at
  • Subject: [mg62711] Re: [TS 28968] Mathematica SingularValueDecomposition failure
  • From: John Sidles <sidles at>
  • Date: Fri, 2 Dec 2005 05:53:09 -0500 (EST)
  • References: <200511111830.jABIUB9M018200@localhost.localdomain>
  • Sender: owner-wri-mathgroup at

Dear Mathematica support (and MathGroup colleagues)

It is now six weeks since my request [TS 28968] asked for help
regarding outright failures in SingularValueDecomposition[]
and routines deriving from it.

Although Wolfram Research always replies promptly, politely,
and respectfully to my emails, the amount of actual support
received to date continues to be zero.

After six weeks with no reply, I am sure you can appreciate
my apprehension that this issue is steadily slipping lower and
lower on Wolfram's internal priority list, such that the waiting
time is asymptotically tending to .... \[Infinity].  Yikes !!

My original questions were:

   (a) Has Wolfram Research reproduced the SVD bug?
   (b) Do you know a workaround at present?
   (c) If no workaround is known, when will one be available?

After conversations with colleagues, there is a new question:

   (d) If the bug turns out to be in LAPACK and/or BLAS
       (as several of my colleagues now suspect) and Wolfram
       Research finds and fixes it, is it then Wolfram Research
       policy to post a correction to the public-domain code?

This is a serious bug that is obstructing a large research
program.  And due to the ubiquity of SVD calculations, it
impacts a considerable portion of Mathematica's customers.

Thank you for your (still hoped for) customer support.

       John Sidles, Professor
       UW School of Medicine
       UW Quantum System Engineering Group
       ( )

  • Prev by Date: Re: Re: Output display of exponential function in Mathematica
  • Next by Date: Re: Programmatic Search and Replace
  • Previous by thread: Fixed points for system using PlotVectorField
  • Next by thread: Re: Programmatic Search and Replace