Hello Friends,
I am trying to solve a optimization problem using CVX. I am using exhaustive search, i.e., changing a parameter named “t”. I am aiming to solve the problem for each “t”, then find the optimum solution. The problem is solved for a specific “t”, but for upper than a specific “t” (1024), the following error is produced:
Error using cvx/plus (line 83)
Disciplined convex programming error:
Illegal operation: {invalid} - {real affine}
The word in the bracket saying “invalid” is strange while the word in the second line saying “real affine” makes sense. Following constants are responsible for the error:
2^t - real( a*( 1 + b*( h.'*(X1 + X2)*conj(h) ) ) - c*h.'*X2*conj(h) ) >= 0;
2^t - real( d*( 1 + e*( h.'*(X1 + X2)*conj(h) ) ) - f*h.'*X1*conj(h) ) == 0;
The only variables are X1 and X2 which are hermitian matrices, “t” is a number, h are vectors and other letters are constants. The constraints seem to fit into the rule set of convex programming (both are affine). However, I cannot find out why this error shows up. Thank you in advance.