# Question:Huge MATLAB/Maple Discepency

## Question:Huge MATLAB/Maple Discepency

Maple

First of all, is there a way I could add maple and matlab scirpts on this forum to make my questions more readable ?? I know we can enter latex scripts on a lot of the latex forums and it's very convenient.

---------------------------------------------------------------------------------------

I have a matrix that's a function of a few parameters.  I evaluate this matrix at specified values of those parameters and Maple gives me:

-10123.5740   14316.89761    5065.657044

14316.89761  -20246.81544   3581.960447

5065.657044  3581.960447    0.335338

While Matlab gives me: (x10^4)

-1.012357539219784   1.431689761935272   0.506565704398669
1.621994295553999  -2.024681544639569   0.358196044696839
0.506565704398669   0.358196044696839   0.000033533800000

Already there's some discrepency: element (2,2) ends in 44 in Maple and 45 in Matlab.

1: Does anyone know where I can find out which algorithm Maple and Matlab use for rational function evaluation ?? I'd like to know which one is more accurate.

2: Can I get Maple (or Matlab) to use a better algorithm for rational function evaluation ?? Or truncate less digits during intermediate steps ??

--------------------------------------------------------

Ok so I can live with the matrix being different.  Now for the eigenvalues:

Maple:

-30370.5026083046541
6204.36103895217002
-6203.91393264752242

Matlab:

1.0e+004 *

-3.124157054733836
0.665773406969565
-0.578621902295085

------------------------------------------------------

This I cannot live with.  And neither can the astronaut on his way to Mars taking  my calculations for granted.

------------------------------------------------------

Matlab's eigenvalues match pretty well with the eigenvalues that my FORTRAN code spits out (my FORTRAN code uses Jacobi rotations, which can theoretically be arbitrarily accurate given enough iterations).

How do I find out what algorithm Maple (and Matlab) use for eigenvalues.  And how do I get it to use a more accurate algorithm ?? Or truncate less digits during intermediate steps ????

Any help is GREATLY appreciated

﻿