OK, it means you have to calculate the trace of all the submatrix internal to your main matrix. For instance if your matrix is 5x5:
1 3 7 2 4
9 1 5 3 6
5 4 9 1 7
2 1 5 4 3
0 1 9 7 4
Then the submatrices are:
1...1 3...1 3 7...1 3 7 2...1 3 7 2 4
.....9 1...9 1 5...9 1 5 3...9 1 5 3 6
............5 4 9...5 4 9 1...5 4 9 1 7
......................2 1 5 4...2 1 5 4 3
...................................0 1 9 7 4
(I used the dot instead the space in order to align the figures)
From which the traces are 1, 2, 11, 15, 19
I wrote the following code (in scilab, very similar to matlab):
u=100; // this is the maximum index of the main matrix
k=grand(u,u,'def'); // Generates the main matrix (random numbers)
//-----------------------------------------------//
g=inv(conj(k))'*k; // Calculates the new matrix
for j=1:u,
M=g(1:j,1:j); // Extract the submatrix jxj (probably a more clever method exists)
z(j)=log10(1+trace(M)); // Calculate the trace
end
plot(z)
The first part is the generation of the matrix, just to see if the code works, of course you have to replace it with your real matrix.