math.PR

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

Abstract: Given any connected graph $G$ and any $\delta\in (0,\infty).$ Note edge weight function at the $(k+1)$-th step of $\delta$ once-reinforced random walk ($\delta$-ORRW) $(X_n)_{n\geq 0}$ on $G$ is given by
$w_k(e)=1+(\delta-1)\cdot I_{\{N(e,k)>0 \}}=I_{\{N(e,k)=0 \}}+\delta\cdot I_{\{N(e,k)>0 \}},$ where $N(e,k)$ is the number of times that edge $e$ has been traversed by the walk before time $k$. In this paper, for $\delta$-ORRW on finite connected graph $G$, we mainly verify a large deviation principle (LDP) for its empirical measures, and characterize critical exponent for exponential integrability of a class of stopping times including cover time.
To prove the LDP, through introducing a class of functionals of the empirical measure processes and solving a novel dynamic programming equation associated with these functionals by a variational representation, we prove a variational formula of the limit of the functionals, which plays a key role in completing the proof of the LDP. We also prove that rate function $I_\delta$ of the LDP is continuous and decreasing in $\delta\in [1,\infty)$, and is constant in $\delta\in (0,1],$ and is not differentiable at $\delta=1$, and $I_{\delta_1}\not=I_{\delta_2}$ for any $1\leq\delta_1<\delta_2.$ For the critical exponent, we prove that it is continuous and strictly decreasing in $\delta\in (0,\infty)$, and there is no exponential integrability at it; and its limit as $\delta\to 0$ can reflect structure of the graph $G$ to a certain extent.
 Subjects: Probability (math.PR) Cite as: arXiv:2206.12801 [math.PR] (or arXiv:2206.12801v1 [math.PR] for this version)

## Submission history

From: Xiangyu Huang [view email]
[v1] Sun, 26 Jun 2022 06:44:48 GMT (488kb)

Link back to: arXiv, form interface, contact.