Escape rate of Markov chains on infinite graphs (Q2248941): Difference between revisions

A connected graph without loops and with an infinite set of vertexes \(V\) and a set of edges \(E\) is called weighted if it is equipped with two weight functions: a measure \(\mu\) on \(V\) and a symmetric function \(w=w(x,y)>0,\;(x,y)\in E\subseteqq V\times V,\;x\neq y.\) It is known that with such weighted graph a symmetric (= time reversible) Markov chain is naturally associated. A common way of constructing a random walk on the weighted graph with \(Q\)-matrix \(Q=\{q_{x,y}\}\) is to set \[ q_{x,y}=\frac{w(x,y)}{\mu(x)},\quad x\neq y \] and \[ q_{x,x}=\frac{1}{\mu(x)}\sum_{y\in V}w(x,y):= \mathrm{deg}(x), \] where \(\mathrm{deg}(x)\) is the degree of \(x\in V.\) Then, a minimal càdlàg Markov chain \((X_t)_{t\geq 0}\) on the graph corresponding to the \(Q\)-matrix can be explicitly constructed. The paper is devoted to the study of the upper rate function for the process \((X_t)_{t\geq 0}\) with respect to a given metric \(d\) on \(V\). The rate function describes the escape rate of the process \((X_t)\).