Error bounds for convex differentiable inequality systems in Banach spaces (Q2576731): Difference between revisions

Consider \((X,\left\| \cdot\right\| )\) a normed vector space and \(f_{i}:X\rightarrow{\mathbb R},\) \(1\leq i\leq k,\) continuous convex functions. Let \(S\) be the set of those \(x\in X\) verifying \(f_{i}(x)\leq 0\) for \(1\leq i\leq k,\) assumed to be nonempty. One says that the Hoffman error bound holds for the preceding system if \(d(x,S)\leq\tau\sum_{i=1}^{k}[f(x)]_{+}\) for all \(x\in X\), for some \(\tau>0;\) when this inequality is satisfied for all \(x\in X\) with \(\left\| x-x_{0}\right\| \leq\varepsilon\) for some \(x_{0} \in\text{ Bd}S\) and some \(\varepsilon>0,\) on says that the Hoffman error bound holds at \(x_{0}.\) First it is shown that when the Hoffman error bound holds at \(x_{0},\) the Abadie qualification condition is satisfied at any \(x\in S\) sufficiently close to \(x_{0},\) and the converse is true when the functions \(f_{i}\) are Gâteaux differentiable in a neighborhood of \(x_{0}.\) Then the authors concentrate on the study of the problem of Hoffman error bound for the case of convex /affine inequalities/equalities. Finally, the authors obtain a Hölder global error bound result.