ABSTRACT
In the previous chapter, the KKT optimality conditions was studied using the normal cone as one of the main vehicles of expressing the optimality conditions. One of the central issues in the previous chapter was the computation of the normal cone at the point of the feasible set C where the set C was explicitly described by the inequality constraints. In this chapter our approach to the KKT optimality condition will take us deeper into convex optimization theory and also we can avoid the explicit computation of the normal cone. This approach uses the saddle point condition of the Lagrangian function associated with (CP ). We motivate the issue using two-person-zero-sum games.
