Enhanced Karush-Kuhn-Tucker condition and weaker constraint qualifications (Q353154): Difference between revisions

The enhanced Fritz-John and KKT optimality conditions are derived for a nonsmooth finite-dimensional optimization problem with equality, inequality and set constraints using the limiting subdifferential of Mordukhovich. Here, instead of the complementary slackness condition (CS), the stronger complementarity violation condition (CV) of Bertsekas with additional subdifferentiation assumptions is taken into consideration. The stronger CV allows to set weaker constraint qualifications (CQ) for the validity of the KKT-conditions. Here, some generalized notions of no nonzero abnormal multiplier CQ (NNAMCQ), pseudonormality and quasinormality of a feasible point are used. It is shown that the pseudonormality or quasinormality implies the calmness of the constraint set valued function which yields a local error bound condition (Ljusternik therorem) and hence a kind of local exact penalty for the problem in consideration. The application of the derived enhanced optimality conditions under quasinormality to the value function of sensitivity analysis gives tighter bounds for its subdifferential. Most of the used tools are introduced in the preliminaries and all new proves are given in detail.