Generalized differential properties of the Auslender gap function for variational inequalities (Q2483997)

Let \(X\) be a finite-dimensional Euclidean space, \(F: X \to X\) a continuously differentiable mapping, and \(K\) a nonempty closed compact subset of \(X\). The authors study the gap function \(\phi(x)=\max \{ \langle F(x), x-y \rangle: y \in K \}\) associated with the variational inequality problem \(\langle F(x^*), x^*-x \rangle \leq 0\quad \forall x \in K\). Given a function \(V: X \to \mathbb R \cup \{ \infty\}\), define the lower contingent derivative \(D_{\downarrow}V({\widehat x})(u)=\limsup_{h \downarrow 0, z \to u}[V({\widehat x}+h z)-V({\hat x})]/h\) \(({\widehat x} \in \text{{dom}}V)\). It is shown that, under suitable conditions, \(D_{\downarrow} \phi({\widehat x})(u)=\langle \nabla F({\widehat x})u, {\widehat x}-{\overline x} \rangle+\sup (\bigcup_{y \in T(K,{\overline x})} \langle F({\widehat x}), u- y\rangle )\), where \(\phi({\widehat x})=\langle F({\widehat x}), {\widehat x}-{\overline x}\rangle\) \(({\overline x} \in K)\) and \(T(K,{\overline x})\) is the Bouligand contingent cone of \(K\) at \({\overline x}\).