scientific article; zbMATH DE number 1487990

From MaRDI portal
Publication:4495119





This article is a valuable contribution to the topic of regularity properties, which theoretically prepares numerical methods for optimization theory in future. As the considered set-valued function \(M\), mapping a Hausdorff real topological space \(E\) into the power set of a real normed vector space \(F\), is, e.g., a parametric feasible set or part of an optimality condition, their regular behaviour may ensure convergence, stability properties and error bounds. The article consists of three sections. Let some closed \(S\subset E\), \(\overline x\in S\), \((\overline x,\overline y)\in gph M\) and an extended real-valued \(f\) on \(E\), finite at \(\overline x\), be given. Section 1 recalls the contingent cone, the Clarke tangent cone to \(S\) at \(\overline x\), in case of coincidence, tangential regularity of \(S\) at \(\overline x\), and, furthermore, directional regularity of \(f\) at \(\overline x\), whereby both generalized and Hadamard directional derivative coincide. By definition, the graph of an \(R\)-derivative of \(M\) at \((\overline x,\overline y)\) is the \(R\)-(contingent or Clarke tangent) cone to \(gph M\) at \((\overline x,\overline y)\). Former result of the authors say. At \((\overline x,\overline y)\), directional regularity of \(\Delta_M(x,y): =d(y,M(x))\) implies tangential regularity of \(gph M\) (the converse is true for finite \(\dim(F))\) and, in some setting under Robinson's constraint qualification, it characterizes tangential regularity of the implicit \(M\). The new results from Section 2 are concerned with \(M\) being Lipschitz epigraphic around \((\overline x,\overline y)\) which becomes characterized by splitting \(F\) into a direct sum, \(f(x,y)\) being upper semicontinuous, partially Lipschitzian and, with the help of its epigraph, \(M(x)\) represented locally. If, moreover, \(M\) is lower semicontinuous at \(\overline x\), then it is tangentially regular at \((\overline x,\overline y)\) iff \(\Delta_M\) is directionally regular (in some equivalent norm on \(F)\). Section 3 draws the attention to tangential regularity of images, assuming \(E,F\) to be normed vector spaces and \(M\) to be pseudo-Lipschitz at \((\overline x,\overline y)\). Now, tangential regularity of \(M\) at \((\overline x,\overline y)\) implies tangential regularity of \(M(\overline x)\) at \(\overline y\). In all of its various parts, this deep analysis is clearly written and carefully proved. Practical utilization seems to be promising. For the entire collection see [Zbl 0935.00054].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4495119)