When invariance implies exchangeability (and applications to invariant Keisler measures)
Samuel Braunfeld,Colin Jahel,Paolo Marimon
2024-08-16
Abstract:We study the problem of when, given a homogeneous structure $M$ and a space $S$ of expansions of $M$, every $\mathrm{Aut}(M)$-invariant probability measure on $S$ is exchangeable (i.e. $S_\infty$-invariant). We define a condition of $k$-overlap closedness on $M$ which implies exchangeability of random expansions by a class of structures whose finite substructures are all contained in a hereditary class with labelled growth-rate $O(e^{n^{k+\delta}})$ for all $\delta>0$. For example, the generic tetrahedron-free $3$-hypergraph is $2$-overlap closed and so all of its random expansions by graphs are exchangeable. Our more general results on $k$-overlap closedness for homogeneous structures imply the same for every finitely bounded homogeneous $3$-hypergraph with free amalgamation. Our results extend and recover both the work of Angel, Kechris and Lyons on invariant random orderings of homogeneous structures and some of the work of Crane and Towsner, and Ackerman on relatively exchangeable structures. We also study conditions under which there are non-exchangeable invariant random expansions looking at the universal homogeneous kay-graphs.
In the second part of the paper we apply our results to study invariant Keisler measures in homogeneous structures. We prove that invariant Keisler measures can be viewed as a particular kind of invariant random expansion. Thus, we describe the spaces of invariant Keisler measures of various homogeneous structures, obtaining the first results of this kind since the work of Albert and Ensley. We also show there are $2^{\aleph_0}$ supersimple homogeneous ternary structures for which there are non-forking formulas which are universally measure zero.
Logic,Combinatorics,Dynamical Systems,Probability