Characterizing quasiconvexity of the pointwise infimum of a family of arbitrary translations of quasiconvex functions, with applications to sums and quasiconvex optimization (Q2230943)

An essential goal of this paper is to find sufficient conditions or even characterizations for quasiconvex functions such that sum or minimum of two (or finitely many) such funtions are again quasiconvex. To do this, the authors use the connection between quasiconvex functions \(f\) and quasimonotone operators (think of \(\partial{f}\)), use a new notion of quasiconvex pairs, and, of course, start with a discussion of the existing results towards the above goal of the paper. Let \(X\) be a Banach space. A family \(A\) of two operators \(T_1,T_2 : X \rightarrow X^*\), will be called a quasimonotone pair, if the operator \(T\) with graph \(\operatorname{gr} T = \operatorname{gr} T_1 \cup \operatorname{gr} T_2\) is quasimonotone. A family of lsc-functions \(f_i,i\in I=\{1,2\}: X \rightarrow R\cup{+\infty}\), is called a quasiconvex pair if for every \(i,j\in I\), every \(x, y\in X\), \(z \in ]x, y[\), \(f_i(x) < f_i(z) \rightarrow f_j(z) \leq f_j(y)\) holds. With that notion of pairs, the authors prove \( \{f_1, f_2\}\) is a quasiconvex pair if and only if \(\{\partial f_1, \partial f_2\}\) is a quasimonotone pair. A lot of properties of pairs follow. For instance Proposition 1 in Chapter 3 says: if \(\{ f_1, f_2\}\) is a quasiconvex pair of lsc-functions, then \(\operatorname{dom} f_1 \cap \operatorname{dom} f_2 \not=\emptyset\). And the next remark is, that without lower semicontinuity of both functions one might get disjointness of the domains. Also remarks are included concerning necessity of Banach space structure in proofs. Chapter 4 deals with sum and minimum of quasiconvex functions. Quasiconvexity is maintained in the case of summation and in the case of minimum formation of quasiconvex pairs (even with more than 2 functions). Some interesting comparisons with already existing results in the literature are added. Chapter 5 embraces the case that the set \(I\) is a convex set of an vector space. Then properties of quasiconvex families are proved. For instance Theorem 9: If \(f_i\), \(i\in I\) is a family of lsc-functions, then it is a quasiconvex family if and only if \(\partial f_i\), \(i\in I\) is a quasimonotone family, or Proposition 6: Let \(f_i\), \(i\in I\) be a quasiconvex family of lsc functions. If at least one of these functions has a lower level set that is weakly compact, then \(\bigcap_{i\in I} \operatorname{dom} f_i \not= \emptyset\). Chapter 6 contains a characterization of the convexity of the union of convex sets. Chapter 7 contains applications in quasiconvex optimization and concerns (extending results from the literature) e. g. conditions for the convexity of the (Benson) proper efficiency solution set of a quasiconvex vector optimization problem (in locally convex spaces) and conditions for a zero duality gap (w.r.t. Lagrangian duality and for a problem class in finite dimensional spaces).