No label defined (Q4938944)

Consider a random variable \(X\) on a metric space \(M\); in the literature, the barycenter of \(X\) is often defined as a subset of \(M\); in particular, different definitions are due to Doss, Herer, and Emery-Mokobodzki. The aim of this work is to derive from each of these definitions another barycenter which will be a point of \(M\). Of course, this can be done only under some assumptions on \(M\) (convexity, negative curvature). First, the authors define the barycenter of a finite number of points; then they deduce the general case by taking the limit with a reverse martingale argument. They also verify that their barycenter satisfies the strong law of large numbers and the ergodic theorem. For the entire collection see [Zbl 0924.00016].