Optimality conditions for an isolated minimum of order two in \(C1\) constrained optimization (Q1022996): Difference between revisions

The article deals with first-order and second-order optimality conditions for an isolated local minimizer of order two of a finite-dimensional optimization problem with an open set constraint and finitely many inequality constraints. First, second-order primal and dual necessary optimality conditions are expressed by critical directions, and it is assumed for them that the functions involved in the problem are once continuously differentiable and that a certain subset of these is second-order directionally differentiable in any critical direction at the minimizer. Next, sufficient optimality conditions of Karush-Kuhn-Tucker type are proven under generalized convexity assumptions, and the concept of Kuhn-Tucker-invexity is studied which provides a necessary and sufficient condition for a global minimizer of the problem. In order to receive a similar characterization of an isolated global minimizer of order two, the concept of Kuhn-Tucker-invexity is extended to so-called strong Kuhn-Tucker-invexity. At last it is shown that strong invexity is a generalization of strong or uniform convexity respectively.