Stability kernel of a multicriteria optimization problem under perturbations of input data of the vector criterion (Q2044064)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Stability kernel of a multicriteria optimization problem under perturbations of input data of the vector criterion |
scientific article |
Statements
Stability kernel of a multicriteria optimization problem under perturbations of input data of the vector criterion (English)
0 references
4 August 2021
0 references
This paper is a contribution to a series of papers of the authors and coauthors w.r.t. stability properties of optimization problems. Here multicriteria optimization problems \[ Q(F,X):\max\{F(x)|x\in X \},\, X \subseteq \mathbb{R}^n,\,F(x)=(f_1(x),\ldots,f_l(x)),\, l\geq 2,\, X \not= \emptyset, \] are considered with \(f_i\), \(i=1,\ldots,l\), continuous, \(P(F,X)\) is the set of Pareto optimal solutions of \(Q(F,X)\), \(Sl(F,X)\) is the set of Slater solutions, \(Sm(F,X)\) is the set of Smale solutions. It is \(Sl(F,X)=\{x\in X | \text{ there is no } y\in X,\, F(y) > F(X)\}\), \(Sm(F,X)=\{x \in X | \text{ there is no } y \not= x,\, F(y) \geq F(X)\}\), and so \[ Sm(F,X) \subset P(F,X) \subset Sl(F,X). \] Then \(Q(F,X)\) is considered with perturbed input data and discussions and theorems follow w.r.t. the influence of that perturbations. Especially stable influence and necessary and sufficient conditions for stability are given. The mentioned perturbations are introduced in the following way: Let \(U \subset \mathbb{R}^m\) the set of input data susceptible to perturbation and a set of perturbed input data \(O_{\delta}(u)= \{u(\delta)\in U | \; ||u(\delta) - u||< \delta \}\). So one has a problem with perturbed input date \[ Q(F_{u(\delta)}, X) : \max Q(F_{u(\delta)}(x)|x \in X), \] where \( u(\delta) \in O_{\delta}(u). \) Using such perturbations a so-called stability kernel \(Ker\) is defined: For one of the considered cases it is \[ Ker(P(F,X)) = \{x\in P(F,X)|\forall \epsilon >0,\, \exists \delta >0,\,\forall u(\delta)\in O_{\delta}(u) \;(x\in O_{\epsilon}(P(F_{u(\delta)},X))\}. \] And then, under conditions for the data of \(Q(F,X)\), the problem \(Q(F,X) \) is stable if \(Ker(P(F,X))=P(F,X).\) Interestingly, with this \((\epsilon, \delta)\)-definition, a lower semicontinuous mapping is existing between \(U\) and \(2^X\). The mentioned sufficient and necessary conditions for stable perturbations are given with help of \(Sl\) and \(Sm\) and are under conditions e.g. of the following shape: closure of \(P(F,X)\) = closure of \(Sm(F,X).\)
0 references
multicriteria optimization problem
0 references
vector criterion
0 references
Pareto-optimal solutions
0 references
Slater set
0 references
Smale set
0 references
perturbations of initial data
0 references
stability
0 references
stability kernel
0 references
0 references
0 references
0 references
0 references
0 references
0 references