Hi, Johan
I tried the following simple code to test equality constraint.
------------------------------------------------------------------------------------------------
P = sdpvar(4,4,'full') % define a 4 by 4 matrix
C = [sum(P,2)==1]; % row sum equals to 1
solvesdp( C, norm(P-eye(4),2)) % minimize the 2-norm P - I
PP = double(P)
------------------------------------------------------------------------------------------------
Matlab resutls:
Result: infeasibility
The LMI constraints were found INFEASIBLE
PP =
NaN NaN NaN NaN
NaN NaN NaN NaN
NaN NaN NaN NaN
NaN NaN NaN NaN
Intuitively, the optimal value P should be identity matrix. If I remove the equality constraint
solvesdp( [], norm(P-eye(4),2)) % minimize the 2-norm P - I
It gives the correct answer, i.e., P = I.
Do I code the equality constraint wrongly? How shall I fix the problem? Thanks in advance.