From: Zinsser <zinsser@...>
Date: 2006-02-02T06:46:08+09:00
Subject: Re: Is this a floating point precision problem?

Todd S. wrote:

> 
>> 
>> I don't get the same inverse; maybe that's a good place to look?
>> 
> 
> That's strange.  I ran it again and this time came up with a different
> number...
> 
> irb> a = Matrix[[-0.0108226964530337, -0.224934679233174,
> -0.974313737622412], [0.0468247818757306, 0.973187905351297,
> -0.225194894880513], [-0.998844486916645, 0.0480592442327619, 0.0]]
> 
> irb> puts a.inverse
> 
> Matrix[[8.0, 2.0, -0.998844486916645], [-0.25, 1.0, 0.0480592442327619],
> [-0.974313737622412, -0.225194894880513, -2.16280922232366e-17]]
> 

I tried it in C++

a:
-0.0108227 -0.224935 -0.974314
0.0468248 0.973188 -0.225195
-0.998844 0.0480592 0

inverse of a:
-0.0108227 0.0468248 -0.998844
-0.224935 0.973188 0.0480592
-0.974314 -0.225195 -7.06138e-16

a * inverse:
1 2.65358e-17 1.1294e-18
-6.53232e-18 1 -3.77082e-18
-8.77526e-19 9.58502e-18 1

Caleb is using a.inv, 
whereas you are using a.inverse.

Perhaps that's your problem ...