The mean-median map (Q2058984)

This is an extended summary of the author's talk ``The Mean-Median Map'' presented in the 2019 conference ``Aperiodic Order Meets Number Theory''. The mean-median map is defined over the space of finite (multi)sets of real numbers, and enlarges such a set by adjoining to it the unique real number for which the mean of the resulting set is equal to the median of the original set. An open conjecture states that, for any finite initial set, the sequence of the numbers adjoined to the set as the map is iterated is eventually constant. In the simplest non-trivial case, i.e., that of three-element initial sets, every initial set is conjugate to \([0,x,1]\) for some \(x\in[1/2,2/3]\), and \textit{F. Cellarosi} and \textit{S. Munday} [J. Difference Equ. Appl. 22, No. 3, 428--440 (2016; Zbl 1382.37039)] have presented a computer-assisted proof that a set of this form obeys the conjecture in the cases of \(x\) lying in small neighbourhoods of all rationals with denominator at most 18. In the present work, the map is redefined over the space of finite sets of piecewise-affine continuous functions with rational coefficients, so that a functional theory of the map can be developed, which results in an extension of the validity of the conjecture to neighbourhoods of 2791 rationals. The significant increase in the number of these rationals turns out to result only in an insignificant increase in the total Lebesgue measure of the neighbourhoods, suggesting the existence of an inaccessible region with presently unexplained dynamical behaviour, captured by a final conjecture that the Hausdorff dimension of the graph of the limit function of the set \([0,x,1]\) is greater than one (see [\textit{J. Hoseana} and \textit{F. Vivaldi}, J. Comput. Dyn. 7, No. 1, 83--121 (2020; Zbl 1448.37024)]). For the entire collection see [Zbl 1459.37002].