Strong stability of stationary solutions and Karush-Kuhn-Tucker points in nonlinear optimization (Q1813594): Difference between revisions

The authors give a survey of different characterizations of strongly stable stationary solutions and strongly regular Karush-Kuhn-Tucker points (KKT) of nonlinear programming problems. First, nonlinear programs with \(C^ 1\) data are considered. It is proved that, for a sufficiently rich class of perturbations, the strong stability of a stationary solution in the sense of \textit{M. Kojima} [in: ``Analysis and computation of fixed points'', Proc. Symp., Univ. Wis. 1979, 93-138 (1980; Zbl 0478.90062)] together with the linear independence constraint qualification are equivalent to the strong stability of the corresponding KKT point (this last property includes also the stability of Lagrange multipliers). Next, for programs with \(C^ 2\) data, it is shown that the strong stability of a KKT point is equivalent to the strong regularity in the sense of \textit{S. M. Robinson} [Math. Oper. Res. 5, 43-62 (1980; Zbl 0437.90094)]. Several other conditions equivalent to the two-mentioned properties are also given, including the nonsingularity of the generalized Jacobian in the sense of \textit{F. H. Clarke} [``Optimization and nonsmooth analysis'' (1983; Zbl 0582.49001)] of some nonsmooth mapping associated with the given program.