Optimality conditions and the basic constraint qualification for quasiconvex programming (Q622397): Difference between revisions
From MaRDI portal
Removed claim: reviewed by (P1447): Item:Q242552 |
Changed an Item |
||
Property / reviewed by | |||
Property / reviewed by: Juan-Enrique Martinez-Legaz / rank | |||
Normal rank |
Revision as of 22:46, 11 February 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Optimality conditions and the basic constraint qualification for quasiconvex programming |
scientific article |
Statements
Optimality conditions and the basic constraint qualification for quasiconvex programming (English)
0 references
31 January 2011
0 references
As introduced in an earlier paper by the authors [J. Optim. Theory Appl. 149, No. 3, 554--563 (2011; Zbl 1229.90208)], a generator of an extended real-valued lsc quasiconvex function \(f\) on a locally convex Hausdorff topological vector space \(X\) is a set of pairs \((k_{i},w_{i})\), where \(k_{i}:\mathbb{R\longrightarrow}\overline{\mathbb{R}}\) is non-decreasing and lsc and \(w_{i}\in X^{\ast }\), such that \(f=\sup_{i\in I}k_{i}\circ w_{i}\). According to Corollary 2.34 in the reviewer's article [Optimization 19, No. 5, 603--652 (1988; Zbl 0671.49015)] (a result referred to as Penot and Volle's theorem in the paper under review) every lsc quasiconvex function admits a generator. The authors define a new notion of subdifferential \(\partial _{G}\), with respect to a generator \(G\), for a lsc quasiconvex function \(f\) on a locally convex Hausdorff topological vector space, and prove that, under mild assumptions, a necessary condition for a point \(x_{0}\) to be a local minimizer of \(f\) over a set \(A\) is \(0\in \partial f(x_{0})+N_{A}(x_{0})\), where \(N_{A}(x_{0})\) denotes the normal cone to \(A\) at \(x_{0}\). They also introduce a new constraint qualification for inequality constrained quasiconvex problems, under which a Kuhn-Tucker type theorem involving the subdifferential of the objective function with respect to \(G\) holds. It turns out that the new constraint qualification is the weakest possible for this theorem. The authors present two examples to show the usefulness of their results.
0 references
optimality condition
0 references
constraint qualification
0 references
quasiconvex programming
0 references
subdifferential
0 references