Invariance principle for variable speed random walks on trees

Abstract

We consider stochastic processes on complete, locally compact tree-like metric spaces $(T,r)$ on their "natural scale" with boundedly finite speed measure $\nu$. Given a triple $(T,r,\nu)$ such a speed-$\nu$ motion on $(T,r)$ can be characterized as the unique strong Markov process which if restricted to compact subtrees satisfies for all $x,y\in T$ and all positive, bounded measurable $f$, \[ \mathbb{E}^x [ \int^{\tau_y}_0\mathrm{d}s\, f(X_s) ] = 2\int_T\nu(\mathrm{d}z)\, r(y,c(x,y,z))f(z) < \infty, \] where $c(x,y,z)$ denotes the branch point generated by $x,y,z$. If $(T,r)$ is a discrete tree, $X$ is a continuous time nearest neighbor random walk which jumps from $v$ to $v'\sim v$ at rate $\tfrac{1}{2}\cdot (\nu(\{v\})\cdot r(v,v'))^{-1}$. If $(T,r)$ is path-connected, $X$ has continuous paths and equals the $\nu$-Brownian motion which was recently constructed in [AthreyaEckhoffWinter2013]. In this paper we show that speed-$\nu_n$ motions on $(T_n,r_n)$ converge weakly in path space to the speed-$\nu$ motion on $(T,r)$ provided that the underlying triples of metric measure spaces converge in the Gromov-Hausdorff-vague topology introduced recently in [AthreyaLohrWinter2016].Comment: 45 page

Scaling Limit for the Ant in High‐Dimensional Labyrinths

“…Under certain assumptions, that will be verified in our context (see (8) in [49] for details), the process . T .…”

“…It was recently shown by Croydon, Hambly, and Kumagai in [27] and by Croydon in [26] that the convergence of measured resistance networks imply the convergence of the corresponding processes (see [8] for a related invariance principle). Furthermore, the convergence of spatial measured networks imply the convergence of the corresponding embedded processes (theorem 7.1 in [26]).…”

A functional limit theorem for irregular SDEs

“…Loosely speaking, the embedding works as follows. We first choose a N (m) (recall that m is the starting point in (2)) and a stopping time ρ 1 such that E(ρ 1 ) = 1/N and M ρ 1 d = Y N 1 . Conditionally on {M ρ 1 = y} we choose a N (y) and a stopping time ρ 2 such that E(ρ 2 ) = 1/N and M ρ 1 +ρ 2 d = y + a N (y)X 2 .…”

“…Our approach to generalize Donsker's theorem is essentially different from the one pioneered by Stone in [12] (also see [2] for a recent generalization to tree-valued processes). In that approach the approximating processes are continuous-time Markov processes that do not jump over points in their state spaces (that is, they can be e.g.…”

Boundedly finite measures: separation and convergence by an algebra of functions

“…But they also appear in very recent research as speed measures of Brownian motions on Ê-trees [AEW13,ALW15] or sampling measures for spatial Fleming-Viot processes [GSW15]. Convergence of metric measure spaces with boundedly finite measures has been analysed with a view towards probabilistic applications in [ALW16].…”