Upper Lipschitz behavior of solutions to perturbed \(C^{1,1}\) programs (Q1584004): Difference between revisions

The author studies the parametric programming problem (MP): \[ \underset {x}{\text{Min}} f(t,x)-a^T x\quad\text{ s.t. }g_i(t,x) \leq b_i,\;h_j(t,x)\leq c_j \quad(i=1, \dots,m;j=1, \dots,k) \] where \(t\in T\subseteq \mathbb{R}^r\) and \(p=(a,b,c) \in\mathbb{R}^{n+m +k}\) are parameters and \(f\), \(g_i\), \(h_j\) are all differentiable functions having a locally Lipschitzian derivative. Investigating mostly the local upper Lipschitz (UL) properties of (MP) the author reports on some characterizations similar to those studied earlier by several researchers including \textit{A. L. Dontchev} and \textit{R. T. Rockafellar} [in: A. V. Fiacco (ed.), Mathematical Programming with Data Perturbations, Lect. Notes Pure Appl. Math. 195, 65-82 (1997; Zbl 0081.90146)]. Local UL continuity of stability and critical points is characterized by injectivity conditions on the contigent derivative of the Kojima function [in: \textit{S. M. Robinsen} (ed.), Analysis and Computation of Fixed Points, Academic Press, New York, 1979 93-138 (1980; Zbl 0478.90062)]. The results already obtained by the author and \textit{B. Kummer} [Comput. Optim. Appl. 13, 61-85 (1999; Zbl 1017.90104)] are extended here and the discussion of stationary solutions avoids the assumption of Mangasarian-Fromovitz CQ. The special case of linear constraints is considered and refinements of known results are also obtained. Further some results pertaining to a second-order stability criterion are also presented.