Uniqueness in linear semi-infinite optimization (Q1318234)

This paper is concerned with developing a general theory of uniqueness in linear semi-infinite optimization. Let \(K\) be a compact subset of \(\mathbb{R}^ d\), \(d\geq 1\) and let \(\{u_ i^ j\}\) and \(\{f_ j\}\) be functions in \(C(K)\). Then the following problem is considered: Minimize \(p({\mathbf a})= \sum_{i=1}^ n a_ i p_ i\) subject to the constraints \[ \sum_{i=1}^ n a_ i u_ i^ j(x)\leq f_ j(x), \qquad x\in K, \quad j=1,\dots,m. \] The paper gives a complete characterization of uniqueness concerning the solutions of the above problem for all choices of functions \(\{f_ j\}\) where \((p_ 1,\dots,p_ n)\in \mathbb{R}^ n\) and the functions \(\{u_ i^ j\}\) are fixed. It turns out that requiring uniqueness for all continuous functions is too restrictive in many cases. Therefore the case of smooth functions is also considered where meaningful results are obtained. For the case \(m=1\) the characterization theorems are extended as follows: These functions \(\{u_ i^ j\}\) are characterized where the above problem has a unique solution for all choices of \(f\in C(K)\) (or \(C^ 1(K)\)) and all \(\{p_ 1,\dots, p_ n\}\) which are contained in some set \(P\subset \mathbb{R}^ n\) ensuring existence of the solution of the optimization problem. The relationship between one-sided \(L_ 1\)- approximation and semi-infinite optimization is used in order to prove these results.