In my CVX code, there is a constraint x < Ek, where x is a variable and Ek is a contant. The problem is feasible and we can always get the optmized value. However, when I gradually enlarge the value of Ek, the obtained objective value sometimes is lower than the original case. What goes wrong? please help me.
Are you maximizing? If not, you don’t seem to have much basis for “complaining”.
Can you show a complete reproducible example, including all solver and CVX output? Perhaps the input data scaling is bad? Does Ek get so big that is causes difficulties or inaccuracy in the solver? Is the difference in optimal objective value “significant”, or just “roundoff” level?