Set-valued means (Q2194608)

The author defines set-valued means \(M\) on nonempty convex subsets \(D\) of a vector space \(X\) as \(M:D^n\rightarrow S(D)\) where \(S(D)\) is the collection of all nonempty subsets of \(D\) and \(M(x_1, \ldots ,x_n)\subset \mathrm{conv}\left\{x_1,\ldots ,x_n\right\}\) and gives a comprehensive guide to their properties. In particular, the notions of reflexive, symmetric and homogeneous are defined analogously to standard means and for convex/concave inequalities are replaced by set containment. Next extensions of the arithmetic mean and quasi-arithmetic means to the set-valued settings are defined (using inverse images instead of inverse functions in the latter case). It is shown that many of the properties of the traditional means have analogous properties for their set-valued counterparts. In order for continuity to translate, \(X\) is assumed to be Hausdorff. The paper concludes with the generalization to Lagrangian-type set-valued means of two real variables using integrals of set-valued functions in the sense of \textit{R. J. Aumann} [J. Math. Anal. Appl. 12, 1--12 (1965; Zbl 0163.06301)].