Stablity of Set-Valued Mappings In Infinite Dimensions: Point Criteria and Applications (Q3128461)

Let \(X\) and \(Y\) be Asplund spaces, and let \(\Phi\) be a subset of the product \(X\times Y\). Use the notation \[ \Phi(x) = \left\{ y\in Y \colon (x,y)\in\Phi \right\} \] to view \(\Phi\) as the graph of a set-valued mapping. The main result in this paper establishes conditions under which the following assumption implies that this mapping has desirable covering properties near a point \((\overline x,\overline y)\) in \(\Phi\): \[ N(\overline x,\overline y;\Phi) \cap \left[\{0\}\times Y^*\right] = \{(0,0)\}. \tag \(*\) \] Here \(N(z;K)\) denotes the well-known sequential normal cone to the set \(K\) at the point \(z\in K\), whose construction and history are outlined in the paper. Condition \((*)\) underlies the title's emphasis on point criteria: in contrast with the neighbourhood criteria used by earlier authors, in which the properties of all sets \(N(x,y;\Phi)\) for \((x,y)\) in \(\Phi\) near \((\overline x,\overline y)\) came into consideration, the current work concentrates the entire burden of sufficiency exactly on the point of interest. The success of this program relies on working in Asplund spaces instead of general Banach spaces, and on augmenting \((*)\) with a suitable compactness hypothesis. This condition, ``partial normal compactness'', is introduced in this paper, and used to obtain the strongest conclusions. Assuming all the hypotheses listed above, the paper's main result asserts that \((*)\) implies that there exist neighbourhoods \(U\) of \(\overline x\) and \(V\) of \(\overline y\), together with a ``covering constant'' \(a>0\) such that \[ \Phi(x+rB) \supseteq \Phi(x)\cap V + a r B \] whenever \(x\) in \(X\) and \(r>0\) are such that \(x+rB \subseteq U\). (Here \(B\) denotes the closed unit ball in the Banach space defined by the context.) This result has immediate consequences for metric regularity and Lipschitzian stability. The authors describe these, and also discuss applications to the sensitivity analysis of infinite-dimensional constraint systems and parametric variational inequalities.