How does CVX run min-max optimization?

I have a worst-case optimization problem. It includes 3 inequality constraints with 3 dual variables lambda1, lambda2 & lambda3. For the first inequality for some given parameters CVX gives the labmda1=1, which is expected. But lambda2 and lambda3 are solved as some unexpected values (7E-7 and 6E-7 respectively) which don’t meet optimality conditions and I don’t know how these values are calculated in CVX. thanks

It’s hard for anyone to say much which is very specific to your problem without your providing details/showing exactly what you did and the results you obtained…

7E-7 and 6E-7 might perhaps be essentially zero, within tolerance. Are optimality conditions satisfied within tolerance, even though not exactly?

What do CVX and the solver report for solution status? Are you getting dual values directly from CVX, or are you calculating them yourself somehow?

1 Like