A profile π=(x1,.,xk), of length k, in a finite connected graph G is a sequence of vertices of G, with repetitions allowed. A median x of π is a vertex for which the sum of the distances from x to the vertices in the profile is minimum. The median function finds the set of all medians of a profile. Medians are important in location theory and consensus theory. A median graph is a graph for which every profile of length 3 has a unique median. Median graphs have been well studied, possess a beautiful structure and arise in many arenas, including ternary algebras, ordered sets and discrete distributed lattices. They have found many applications, for instance in location theory, consensus theory and mathematical biology. Trees and hypercubes are key examples of median graphs. We establish a succinct axiomatic characterization of the median procedure on median graphs, settling a question posed implicitly by McMorris, Mulder and Roberts in 1998 [19]. We show that the median procedure can be characterized on the class of all median graphs with only three simple and intuitively appealing axioms, namely anonymity, betweenness and consistency. Our axiomatization is tight in the sense that each of these three axioms is necessary. We also extend a key result of the same paper, characterizing the median function for profiles of even length on median graphs.

Consensus axiom, Consensus function, Location function, Median, Median function, Median graph
dx.doi.org/10.1016/j.dam.2012.10.027, hdl.handle.net/1765/63998
Discrete Applied Mathematics
Erasmus School of Economics

Mulder, H.M, & Novick, B. (2013). A tight axiomatization of the median procedure on median graphs. Discrete Applied Mathematics, 161(6), 838–846. doi:10.1016/j.dam.2012.10.027