A steepest descent-like method for variable order vector optimization problems (Q467437): Difference between revisions

For vector optimization problems in which the set of points whose image is better depends on the point under consideration, the variable order structure is given by a set-valued application whose image is a proper, pointed, convex, and closed cone, for all \(n\)-dimensional vectors. So, the problem of minimizing a vector function \(F\) with respect to the set-valued order is to find a point such that there does not exist another point with a better (and different) value of the objective function. The authors present the extension of the steepest descent algorithm for classical vector optimization models to the variable order case. This approach appeared in [\textit{J. Fliege} and \textit{B. F. Svaiter}, Math. Methods Oper. Res. 51, No. 3, 479--494 (2000; Zbl 1054.90067)] for solving multicriteria models. It was extended by \textit{L. M. Graña Drummond} and \textit{B. F. Svaiter} [J. Comput. Appl. Math. 175, No. 2, 395--414 (2005; Zbl 1058.90060)] for convex vectorial models, and recently, in [\textit{J. Y. B. Cruz} et al., Nonlinear Anal., Theory Methods Appl., Ser. A, Theory Methods 74, No. 16, 5268--5273 (2011; Zbl 1225.90114)] the convergence for the quasiconvex case was obtained.\newline In the present paper, the authors obtain the properties of the limit points of the sequence generated by a steepest descent-like algorithm for solving vector optimization problems with variable order. Under a convex-like assumption, they claim that the sequence is bounded and all its accumulation points are solutions of the problem.