Robinson's implicit function theorem and its extensions

DOI10.1007/S10107-007-0161-1MaRDI QIDQ959966zbMATH OpenOpenAlexFDO

Authors A. L. Dontchev, R. T. Rockafellar

Publication date 16 December 2008

Published in Mathematical Programming. Series A. Series B (Search for Journal in Brave)

Full work available at URL https://doi.org/10.1007/s10107-007-0161-1

calmness first--order approximations inverse and implicit function theorem Lipschitz modulus semiderivatives variational inequalities

Mathematics Subject Classification ID

Abstract inverse mapping and implicit function theorems involving nonlinear operators (47J07) Set-valued and variational analysis (49J53) Sensitivity, stability, parametric optimization (90C31)

Review text

The authors consider ``generalized equations with parameters such as \[ f(p,x)+F(x)\ni 0,\tag{1} \] with solution map \[ S(p)=\{x|f(p,x)+F(x)\ni 0\}.\tag{2} \] The central object is the investigation of \(S,\) and, in particular, questions about its possible single-valuedness and differentiability. The starting point of this paper is a pioneer work due to Robinson [see \textit{S. M. Robinson}, Math. Oper. Res. 5, 43--62 (1980; Zbl 0437.90094)]. The aim is to put Robinson's ideas in a broader formulation in order to fully bring out and extend his contributions to generalized implicit function theory. Historical background on the classical implicit function theorem and the classical inverse function theorems can be found, for example, in the book by [\textit{S. G. Krantz} and \textit{H. R. Parks}, The implicit function theorem. History, theory and applications. Boston, MA: Birkhäuser (2002; Zbl 1012.58003)]; both the results are stated in terms of the notion of (single-valued) graphical localization of the solution map \(S\) at a point \(\overline p\) for \(\overline x,\) where \((\overline p, \overline x)\) is the point of \(P\times X\) where the assumptions are satisfied. The implicit function theorem, rather than the inverse function theorem, provides the pattern of result leading to a generalized equation (1) with \(F\not\equiv 0.\) In such a context, while single-valuedness of a graphical localization of \(s\) can still be realistic, its differentiability cannot be expected, and, in some sense, it is substituted by other properties, like calmness and Lipschitz continuity. Indeed, both these properties provide estimates of the rate of change of \(s(p);\) moreover, they are connected with differentiability through notions of approximation. Indeed, if \(A\) denotes a function in \({\mathcal L}(P,X),\) differentiability and strict differentiability of \(s\) at \(\overline p\) can be expressed as clm\((s-A;{\overline p})=0\) and \(\text{lip}(s-A;{\overline p})=0.\) Before getting on with the formulation of Robinson's main theorem (Theorem 1.6), the authors mention some results which indicate how the classical inverse function theorem can be complemented, or sharpened, in terms of these concepts. The last part of Section 1 is devoted to careful and deep comments on Robinson's theorem and its corollaries, and, more generally, on the results of the previous mentioned paper. In Section 2 the authors present a result (Theorem 2.1) which carries Robinson's theorem to a setting of generalized equations (1) without any, even rudimentary, form of differentiability required of \(f.\) First, they introduce the definition of the concept of (strictly) first-order approximation (already proposed by Robinson in 1991), which will be used as a replacement for the linearization condition that appears in Theorem 1.6. Moreover, the notion of moduli for ``partial calmness (\(\widehat{\text{clm}}\)), and Lipschitz continuity (\(\widehat{\text{lip}}\)) are required in order to define partial first-order approximations. Their main result of this section is the following extension of Robinson's implicit function theorem: Theorem 2.1: In the generalized equation (1) with its solution mapping \(S\) in (2), let \(\overline p\) and \(\overline x\) be such that \(\overline x\in S(\overline p).\) Assume that {\parindent=6mm \begin{itemize}\item[(a)] \(f(\cdot, \overline x)\) is continuous at \(\overline p\) and \(h\:X\to Y\) is a function having \(h(\overline x)=f(\overline p,\overline x)\) and\break \(\widehat {\text{clm}}_x(f-h;(\overline p,\overline x))<\infty.\) \item[(b)] For the set-valued mapping \(G=h+F,\) with \(G(\overline x)\ni 0,\) the inverse \(G^{-1}\) has a Lipschitz localization \(\sigma\) at \(0\) for \(\overline x\) with \[ {\text{lip}}(\sigma;0)\cdot \widehat{\text{lip}}_x(f-h;(\overline{p},\overline{x}))<1. \] \end{itemize}} Then \(S\) has a single-valued graphical localizaton \(s\) at \(\overline{p}\) for \(\overline{x}\) which is continuous at \(\overline{p}\) and such that for every \[ \kappa >\left(\text{lip}(\sigma;0)^{-1}-\widehat{\text{lip}}_x(f-h;(\overline{p},\overline{x}))\right)^{-1} \] the estimate holds that for all \(p,p'\) in a neighborhood of \(\overline{p},\) \[ \|s(p')-s(p)\|\leq \kappa \|f(p',s(p))-f(p,s(p))\|. \] In Section 3, the main result (Theorem 3.1) demonstrates that if some relatively mild assumptions about the function \(f\) are assumed, then a first-order approximation of the localized solution mapping \(s\) in Theorem 2.1 can be developed. Indeed, it achieves an implicit function theorem which adds to the extension of Robinson's theorem in Theorem 2.1 in such a manner as to obtain full coverage of the classical implicit function theorem. As a consequence, an exact formula for the calmness modulus and for the Lipschitz modulus of the localized solution mapping \(s\) at \(\overline p\) is obtained.

Recommendations

Cites work

Cited in

(56)

This page was built for publication: Robinson's implicit function theorem and its extensions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q959966)