LabVIEW

cancel
Showing results for 
Search instead for 
Did you mean: 

matrix inverse bug

Hi,
 
We are working with labview8 and the matrix inverse (at least the 4 by 4) differs from the result of matlab, is there something that I could be doing wrong?
0 Kudos
Message 1 of 8
(3,812 Views)
Do you have a matrix for which you get the wrong answer? I tried it with a few matrices and I was getting the same answer.
0 Kudos
Message 2 of 8
(3,802 Views)
You are using the "Inverse Matrix" primitive from the linear algebra palette, not the 1/x primitive of the numeric palette, right?
0 Kudos
Message 3 of 8
(3,797 Views)
here... thanks for the quick response!
0 Kudos
Message 4 of 8
(3,767 Views)
Hi macrom,

atleast the MathScript node gives the same result as the attachment shows...

Edited: changed the mathscript command from "a^-1" to "inv(a)" - both give the same result (within DBL precision).


Message Edited by GerdW on 07-03-2008 01:25 PM
Best regards,
GerdW


using LV2016/2019/2021 on Win10/11+cRIO, TestStand2016/2019
0 Kudos
Message 5 of 8
(3,761 Views)
Hi GerdW,

Thanks for the reply.

Im sorry I cant open the vi file, I only have the labview 8 and the file is saved as 8.5

I've tried the inverseMatrix.vi and the mathscript with both a^-1 and inv(a), all three are consistent with each other but still inconsistent with Matlab and Mathematica. A mathematician friend of mine said that it could be a numerical inestability of the algorithm. Has anyone heard of Labview having this problem? It could still be that its Matlab or Mathematica.

Anyway, I decided to use a matrix that works well in all the cases.

Thank you very much!!!




0 Kudos
Message 6 of 8
(3,744 Views)
Well, Mathscript is basically LabVIEW under the hood, so I don't expect them to be different.
 
How different are the results between LabVIEW and matlab?
 
I did some test and the LabVIEW version seems OK.
  1. If you take the inverse twice, you get the input matrix back within 1e-17.
  2. I also tried the scilab script node, the results differ from LabVIEW by about 1e-16, or about the precision of DBL. So clearly they use an entirely different algorithm, but give the same result!

 



Message Edited by altenbach on 07-03-2008 08:13 AM
0 Kudos
Message 7 of 8
(3,729 Views)
Hello macrom,

If it is just a small difference, it is most likely due to algorithmic differences.  Can you post the result matrix you get from your computation in the MATLAB® software?

MATLAB® is a registered trademark of The MathWorks, Inc.

Grant M.
Staff Software Engineer | LabVIEW Math & Signal Processing | National Instruments
0 Kudos
Message 8 of 8
(3,678 Views)