A down-up chain with persistent labels on multifurcating trees

TL;DR

This paper introduces a novel down-up Markov chain model for multifurcating trees with labeled leaves, extending previous binary models and preserving key structural properties over time.

Contribution

It develops a general framework for down-up chains on multifurcating trees, ensuring label dynamics that maintain the alpha-gamma distribution and branch point order.

Findings

Preserves labeled alpha-gamma distribution over time

Maintains branch points between smallest labels for quadratic time

Proposes conjectures on diffusive scaling limits of the process

Abstract

In this paper, we propose to study a general notion of a down-up Markov chain for multifurcating trees with n labelled leaves. We study in detail down-up chains associated with the $(\alpha, \gamma)$-model of Chen et al. (2009), generalising and further developing previous work by Forman et al. (2018, 2020) in the binary special cases. The technique we deploy utilizes the construction of a growth process and a down-up Markov chain on trees with planar structure. Our construction ensures that natural projections of the down-up chain are Markov chains in their own right. We establish label dynamics that at the same time preserve the labelled alpha-gamma distribution and keep the branch points between the k smallest labels for order $n^2$ time steps for all k larger than 2. We conjecture the existence of diffusive scaling limits generalising the "Aldous diffusion" by Forman et al. (2018+) as a continuum-tree-valued process and the "algebraic {\alpha}-Ford tree evolution" by L\"{o}hr et al. (2018+) and by Nussbaumer and Winter (2020) as a process in a space of algebraic trees.