Extremality, stationarity and generalized separation of collections of sets (Q2315263)

Separations of convex or non-convex sets \(M_i\) are in optimization, convex analysis and related fields among others a basis for optimality conditions. Since about 40 years (compare [\textit{S. Dempe} et al., Optimization 56, No. 5--6, 577--604 (2007; Zbl 1172.90481)] or f. i. [\textit{B. S. Mordukhovich}, Variational analysis and generalized differentiation. I: Basic theory. II: Applications. Berlin: Springer (2005; Zbl 1100.49002); \textit{H. T. Bui} and \textit{A. Y. Kruger}, Vietnam J. Math. 46, No. 2, 215--242 (2018; Zbl 1391.49024)]) so-called extremal principles have been used as well. Here (in the neighborhood of an extremal point \(x\) in \(\bigcap {M_i})\) separation is replaced by a more general local geometric extremality property: by sufficiently small translations of the sets \(M_i\) it is achieved that their intersection is empty. Thus, a very large number of statements of both known and significantly extended results concerning esp. w. r. to separations and optimality conditions (and all these things with different parameters and assumptions (e.g., \(\bigcap {M_i}\) empty)) are proven. In this paper, starting with an interesting Introduction, the authors carefully discuss relationships between all the statements mentioned above. In Section 2 esp. the geometric Ekeland variational principle (GEVP) and the asymmetric geometric Ekeland variational principle (AGEVP) are introduced together with the proof of their equivalence to the conventional Ekeland variational principle followed by normal cones, subdifferential sum rules and the extended extremal principle. And, as the authors write, ``more importantly, quantitative estimates for parameters involved in such statements'' are discussed. In Section 3 one finds discussions concerning constraint qualifications, qualification conditions in subdifferential/coderivative calculus and convergence analysis of computational algorithms. Section 4 deals with distance between sets, transversality of a collection of sets and f. i. the role of the nonintersect index. Sections 5 and 6 (using the results of the first sections) contain quite a lot of new properties, relationships and in particular characterizations in connection with extremality, stationarity, transversality and w. r. to primal and dual topics too. There are also considerations to the usefulness of Asplund spaces. Theorem 6.1 (and its corollaries and two following theorems) give dual necessary characterizations of extremality (one finds together primal and dual vectors), and remarks show interesting connections (also equivalences) with f. i. Ekeland's variational principle and with the separation due to [\textit{X. Y. Zheng} and \textit{K. F. Ng}, SIAM J. Optim. 21, No. 3, 886--911 (2011; Zbl 1229.49018)] and also a remark w. r. to extensions to infinite collections of sets is added.