mma v.2.1 's possible bug ?
- To: mathgroup at yoda.physics.unc.edu
- Subject: mma v.2.1 's possible bug ?
- From: yoda at ele.crl.melco.co.jp
- Date: Fri, 13 Nov 92 14:40:16 JST
Dear mma users, I have compared mma v.2.1 with v.2.0 on my Macintosh
IIci(both are enhanced versions).
My understanding is that LinearSolve or Solve for matrix equation solving
under v.2.1
leads to somewhat strange results. Here's a simple example.
---version 2.0 gives a reasonable solution.----
ma=Array[m,{2,2}]
{{m[1, 1], m[1, 2]}, {m[2, 1], m[2, 2]}}
bv=Array[b,2]
{b[1], b[2]}
LinearSolve[ma,bv]
b[2] m[1, 2]
{-(------------------------------------) +
-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2]
b[1] m[2, 2]
------------------------------------,
-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2]
b[2] m[1, 1]
------------------------------------ -
-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2]
b[1] m[2, 1]
------------------------------------}
-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2]
--- version 2.1 gives more complicated solution, which can be still
simplified!!! ---
ma=Array[m,{2,2}]
{{m[1, 1], m[1, 2]}, {m[2, 1], m[2, 2]}}
bv=Array[b,2]
{b[1], b[2]}
LinearSolve[ma,bv]
{(-(m[1, 2] (b[2] m[1, 1] - b[1] m[2, 1])) +
b[1] (-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2])) /
(m[1, 1] (-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2])),
b[2] m[1, 1] - b[1] m[2, 1]
------------------------------------}
-(m[1, 2] m[2, 1]) + m[1, 1] m[2, 2]
----------------------
When I tried to use LinearSolve for larger matrix, the CPU time was about
10 times longer
in v.2.1 than in v.2.0!!!
***************************************
* Kiyoshi Yoda, Ph.D *
* *
* Mitsubishi Electric Corporation *
* Central Research Laboratory *
* 8-1-1 Tsukaguchi-Hommachi *
* Amagasaki, 661 Japan *
* *
* yoda at ele.crl.melco.co.jp *
* domestic phone: 06-497-7130 *
* domestic fax : 06-497-7288 *
* international fax: +81-6-497-7288 *
***************************************