Large deviation principle for empirical measures of once-reinforced random walks on finite graphs

Autor: Huang, Xiangyu, Liu, Yong, Xiang, Kainan
Rok vydání: 2022
Předmět:
Druh dokumentu: Working Paper
Popis: In this paper, we focus on studying the long time behaviors of a type of random walk called the $\delta$ once-reinforced random walk ($\delta$-ORRW) on a finite connected graph $G$ with at least 3 vertices for $\delta>0$. This random walk process involves edge weights and the traversal of edges over time, where the edge weight function at the $(k+1)$-th step is given by \[ w_k(e)=1+(\delta-1)\cdot I_{\{N(e,k)>0 \}} {\rm\ equals\ } 1 \ {\rm if}\ N(e,k)=0,\ {\rm and\ equals\ } \delta \ {\rm if}\ N(e,k)>0, \] and $N(e,k)$ is the number of the traversal of edge $e$ before time $k$. Our main goal is to establish a large deviation principle (LDP) for the empirical measures of this random walk and to determine the critical exponent for the exponential integrability of certain stopping times, including the cover time. To prove the LDP, we introduce a set of functionals related to the empirical measure processes and solve a new dynamic programming equation associated with these functionals using a variational representation. This enables us to derive a variational formula for the limit of these functionals, which is crucial for completing the proof of the LDP. Additionally, we demonstrate that the rate function $I_\delta$ of the LDP is continuous and decreasing in $\delta\in[1,\infty)$, remains constant in $\delta\in(0,1]$, and is not differentiable at $\delta=1$. Furthermore, we show that $I_{\delta_1}\not=I_{\delta_2}$ for any $1\leq\delta_1<\delta_2$. Regarding the critical exponent, we establish its continuity and strict decrease in $\delta\in(0,\infty)$, and we determine that there is no exponential integrability at this critical exponent. We also find that the limit of the critical exponent as $\delta\to0$ is infinite if and only if $G$ is a 3-vertex connected graph or a star-shaped graph.
Databáze: arXiv