Metric aggregation functions revisited

Abstract

In 1981, J. Borsík and J. Doboš analyzed what properties a function must satisfy in order to merge a collection of metric spaces into a single one. Later on, E. Castiñeira, A. Pradera and E. Trillas studied a variant of the same problem in which each metric of the collection to be merged is defined on the same non-empty set. In this, paper we continue the work in this last aforesaid direction. On the one hand, we provide a new characterization of such functions and a few methods to construct them. On the other hand, we analyze the existence of absorbent, idempotent and neutral elements for such class of functions and, thus, we design techniques that allow to discard those functions that are not useful for merging metrics. Finally, we discuss when the functions under study preserve metrics.