Long Time Results for a Weakly Interacting Particle System in Discrete Time

TL;DR

This paper investigates the long-term behavior of weakly interacting particle systems in unbounded spaces, demonstrating convergence properties, fixed point uniqueness, and propagation of chaos with explicit convergence rates.

Contribution

It extends previous results to unbounded state spaces, proving exponential convergence to a unique fixed point and establishing uniform convergence of empirical measures.

Findings

Unique fixed point for the nonlinear Markov evolution.

Exponential convergence to the fixed point from arbitrary initial conditions.

Uniform in time convergence of empirical measures with explicit rates.

Abstract

We study long time behavior of a discrete time weakly interacting particle system, and the corresponding nonlinear Markov process in $\mathbb{R}^d$, described in terms of a general stochastic evolution equation. In a setting where the state space of the particles is compact such questions have been studied in previous works, however for the case of an unbounded state space very few results are available. Under suitable assumptions on the problem data we study several time asymptotic properties of the $N$-particle system and the associated nonlinear Markov chain. In particular we show that the evolution equation for the law of the nonlinear Markov chain has a unique fixed point and starting from an arbitrary initial condition convergence to the fixed point occurs at an exponential rate. The empirical measure $\mu_{n}^{N}$ of the $N$-particles at time $n$ is shown to converge to the law $\mu_{n}$ of the nonlinear Markov process at time $n$, in the Wasserstein-1 distance, in $L^{1}$, as $N\to \infty$, uniformly in $n$. Several consequences of this uniform convergence are studied, including the interchangeability of the limits $n\to \infty$ and $N\to\infty$ and the propagation of chaos property at $n = \infty$. Rate of convergence of $\mu_{n}^{N}$ to $\mu_{n}$ is studied by establishing uniform in time polynomial and exponential probability concentration estimates.