Nonasymptotic necessary conditions for nonsmooth infinite optimization problems (Q1083684)

The author studies in this paper necessary optimality conditions for infinite programming problems of the form: minimize f(x) on \(\{x\in D| g_ t(x)\leq 0\) for each \(t\in T\}\). Here D is a nonempty open subset of a locally convex Hausdorff topological real vector space E and T denotes a compact Hausdorff space. Further, \(f,g_ t,t\in T\), are real valued functionals defined on D that are not assumed to be differentiable or convex. The main results are nonasymptotic conditions of the John, Karush-Kuhn-Tucker type in terms of approximate derivatives, the 'Lagrange multiplier' being a certain nonnegative Radon measure on the index set T. The essential tool for deriving both types of optimality conditions is a quite general Farkas type results for infinite systems of nonlinear inequalities.