Services & ResourcesWolfram ForumsMathGroup Archive

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

Re: Re: SingularValueDecomposition bug: fifty examples

• To: mathgroup at smc.vnet.net
• Subject: [mg61318] Re: [mg61183] Re: SingularValueDecomposition bug: fifty examples
• From: John Sidles <sidles at u.washington.edu>
• Date: Fri, 14 Oct 2005 22:23:09 -0400 (EDT)
• References: <200510060745.DAA08051@smc.vnet.net> <200510120542.BAA09202@smc.vnet.net> <9da7362420fbd510d2708750f0b14fc3@wolfram.com>
• Sender: owner-wri-mathgroup at wolfram.com

On Oct 12, 2005, at 8:01 AM, Bruce Miller wrote:

> General suggestion: if one of you thinks he has an example of
> a bug in Mathematica, please send it to support at wolfram.com.
> ONE example per email, with an explanation of what appears to be
> wrong with that example.
> Include the Mathematica input, the output, and the VERSION NUMBER.
>
> Mr. Sidles's  tarball is being looked at, but it is one
> of many things in that Tech Support engineer's queue.
>
> Bruce

Bruce, you are so right ... Wolfram Research can't help the
Mathematica user community unless and until we provide solid
documentation of reproducible behaviors.

My report number is [TS 28968], and my tarball *does* provide
complete input, output, and version number information (as well
as a full profile of the OS version and the model number of the
machines).

It helps to be patient after submitting a report. E.g.,
the release of SingularValueDecomposition[] in Version 5
provided a workaround to a problem with EigenSystem[] that
was documented in 1998:

   /forums.wolfram.com/mathgroup/archive/1998/Mar/msg00425.html

Wolfram's initial response to the 1998 request was minimalist:
the documentation for Eigensystem[] was updated such that the
problematic behavior became legal.

It wasn't until Version 5 that SingularValueDecomposition[]
provided a reliable workaround -- a six-year wait.

This isn't Wolfram Research's fault. Squashing bugs is a slow
process partly because it is *very* difficult to create reliable
linear algebra packages. LAPACK has been around for decades, and
bugs are still being found (obviously).

Users and engineers have to work together on this. If and when
Technical Support provides a workaround, I'll surely post it.

Meanwhile, best wishes to the engineer(s) assigned this problem!

    Sincerely
    John Sidles, Ph.D.
    Professor, School of Medicine
    Adjunct Professor, College of Engineering

• References:
  • Re: SingularValueDecomposition bug: fifty examples
    • From: John Sidles <sidles@u.washington.edu>