Services & Resources / Wolfram Forums
-----
 /
MathGroup Archive
2007
*January
*February
*March
*April
*May
*June
*July
*August
*September
*October
*Archive Index
*Ask about this page
*Print this page
*Give us feedback
*Sign up for the Wolfram Insider

MathGroup Archive 2007

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

Search the Archive

Re: Options for a very large computation

  • To: mathgroup at smc.vnet.net
  • Subject: [mg77863] Re: Options for a very large computation
  • From: Roman Pearce <rpearcea at gmail.com>
  • Date: Mon, 18 Jun 2007 07:06:35 -0400 (EDT)
  • References: <f531f3$1vp$1@smc.vnet.net>

On Jun 17, 4:12 am, "Bob D" <rdob... at carleton.edu> wrote:
> I'm solving a 24-by-24 linear system symbolically.

This is not advisable.  In general the determinant will have something
like 24! = 6.2 x 10^23 terms, and by Cramer's rule this blowup will
appear in every "solution".  It is unlikely you will even be able to
evaluate the "solutions".  In any case, plugging numbers into any
formula you get will be orders of magnitude more expensive than
plugging numbers into the system and solving it directly.

For what it's worth, there is a way to "symbolically solve" linear
systems, but the output is essentially a program that plugs in numbers
and branches accordingly.  Other people in this newsgroup probably
know more about that than I do.



  • Prev by Date: Re: Help with formatting output
  • Next by Date: Re: Crinkled Hyperbola (ContourPlot in Manipulate)
  • Previous by thread: Options for a very large computation
  • Next by thread: Re: Options for a very large computation