Ranking-based rich-get-richer processes

TL;DR

This paper analyzes a ranking-based Markov process with rich-get-richer dynamics, demonstrating almost sure convergence and characterizing limits, with applications to ranking algorithms and Pólya urns.

Contribution

It introduces a novel framework for ranking-based processes with rich-get-richer effects, providing convergence results and applications to ranking algorithms.

Findings

Almost sure convergence of the process divided by n.

Characterization of possible limit points.

Applicability to ranking-based Pólya urns and algorithms.

Abstract

We study a discrete-time Markov process $X_n\in\mathbb{R}^d$, for which the distribution of the future increments depends only on the relative ranking of its components (descending order by value). We endow the process with a rich-get-richer assumption and show that, together with a finite second moments assumption, it is enough to guarantee almost sure convergence of $X_n$ / $n$. We characterize the possible limits if one is free to choose the initial state, and give a condition under which the initial state is irrelevant. Finally, we show how our framework can account for ranking-based P\'olya urns and can be used to study ranking-algorithms for web interfaces.