The inverse of A(i,j) has signature B(j,i)

GAMS has strict domain checking (type checking). This has some interesting consequences. Basically all good: much better protection against errors compared to simply using integer indices $1,\dots,n$. Consider the square matrix $\color{darkblue}A_{i,j}$:

set   i /a,b,c,d/   j /1,2,3,4/ ; parameter A(i,j) 'square matrix'; A(i,j) = min(ord(i),ord(j)); display A;

The so-called "minij" matrix looks like:

----      7 PARAMETER A  square matrix

            1           2           3           4

a       1.000       1.000       1.000       1.000
b       1.000       2.000       2.000       2.000
c       1.000       2.000       3.000       3.000
d       1.000       2.000       3.000       4.000

The task to invert a parameter $\color{darkred}A_{i,j}$, can be stated as finding a feasible solution to: $$\color{darkred}B\cdot \color{darkblue}A =  \color{darkblue}I$$ or $$\sum_i \color{darkred}B_{j,i}\cdot \color{darkblue}A_{i,j'} = \color{darkblue}I_{j,j'} \> \>\>\forall j,j'$$ The first index of $\color{darkblue}A$ is $i$ which means the second index of $\color{darkblue}B$ is also $i$. 

So our GAMS code should look like:

set   i /a,b,c,d/   j /1,2,3,4/ ; parameter A(i,j) 'square matrix'; A(i,j) = min(ord(i),ord(j)); display A; *-------------------------------------------------- * Invert A by using the identity: *    B*A=I *-------------------------------------------------- alias (j,jj); parameter Ident(j,jj); Ident(j,j) = 1; variable B(j,i) 'inverse of A'; equation Inverse(j,jj); Inverse(j,jj).. sum(i,B(j,i)*A(i,jj)) =e= Ident(j,jj); model m /all/; solve m using cns; display B.l;

This shows:

----     23 VARIABLE B.L  inverse of A

            a           b           c           d

1       2.000      -1.000
2      -1.000       2.000      -1.000
3                  -1.000       2.000      -1.000
4                              -1.000       1.000

The same argument holds when we write: $$\color{darkblue}A \cdot \color{darkred}B =  \color{darkblue}I$$ or $$\sum_j \color{darkblue}A_{i,j} \cdot \color{darkred}B_{j,i'}  = \color{darkblue}I_{i,i'} \> \>\>\forall i,i'$$ The second index of $\color{darkblue}A$ is $j$ which means the first index of $\color{darkblue}B$ is also $j$.

The conclusion is: the inverse of A(i,j) is a symbol with signature B(j,i). 

Note that usually we use a single set $i$ and make $j,k$ aliases. That will not reveal the above issue:

set   i /a,b,c,d/ ; alias(i,j,k); parameter A(i,j) 'square matrix'; A(i,j) = min(ord(i),ord(j)); *-------------------------------------------------- * Invert A by using the identity: *    B*A=I *-------------------------------------------------- parameter Ident(i,j); Ident(i,i) = 1; variable B(i,j) 'inverse of A'; equation Inverse(i,j); Inverse(i,j).. sum(k,B(i,k)*A(k,j)) =e= Ident(i,j); model m /all/; solve m using cns; display A,B.l;

With results:

----     25 PARAMETER A  square matrix

            a           b           c           d

a       1.000       1.000       1.000       1.000
b       1.000       2.000       2.000       2.000
c       1.000       2.000       3.000       3.000
d       1.000       2.000       3.000       4.000


----     25 VARIABLE B.L  inverse of A

            a           b           c           d

a       2.000      -1.000
b      -1.000       2.000      -1.000
c                  -1.000       2.000      -1.000
d                              -1.000       1.000

Similarly, numerical codes for inverting matrices don't worry about this at all: they just see everything in terms of integer index positions $i,j \in \{1,\dots,n\}$.