Jordi / Esther Soler / Bonet

Poor. Sorry, but you should NOT be using this code to find an inverse. If anybody is doing so, they are mistaken.

The fact is, you should almost never be computing the inverse in the first place. The very few times this would be of any value, why not use inv anyway? That would be FAR faster. More stable will be to use pinv, which will not give you trash for nearly singular matrices. Even for a 50x50 system, pinv is just as fast.

In almost all circumstances, use backslash to solve systems of equations. Thus if you wish to solve the system

  A*x = b

Use

  x = A\b

or

  x = pinv(A)*b

Do NOT use invjor here.

There is not much help in this, nor . There are no indications that singularity has been detected. No errors, no warnings are issued. As an example, try this utterly simple problem to solve:

A = [0 1;1 0]
A =
     0 1
     1 0

[U,I] = invjor(A)
Ecuations swapping needed
Ecuations swapping needed
U =
     0 1
     1 0
I =
   Inf NaN
   NaN Inf

Gosh, I could swear the inverse is incorrect here. In fact, A is perfectly well conditioned, and the inverse of A is itself!

A*A
ans =
     1 0
     0 1
 
The author gets the size of the matrix, then proceeds to try to "invert" even a non-square matrix.

A = rand(2,3)
A =
      0.81472 0.12699 0.63236
      0.90579 0.91338 0.09754
>> [U,I] = invjor(A)
??? Attempted to access d(3); index out of bounds because numel(d)=2.

Error in ==> invjor at 43
   if d(j)==1;

>> A = rand(3,2)
A =
       0.2785 0.96489
      0.54688 0.15761
      0.95751 0.97059
>> [U,I] = invjor(A)
??? Attempted to access U(1,3); index out of bounds because size(U)=[3,2].

Error in ==> invjor at 33
         Z(k,i)=U(k,i)/U(i,i); % Multiplier

Sorry, but this is no more than a homework assignment, and not a terribly well done one either. I'd give it a C minus at very best. On the file exchange, it is just an F.

Geert VD's Code is efficient using the companion matrix.
As a rule of thumb with matlab, you should always avoid using the loops if you can use vectorized methods.