Edge-reinforced random walk on a ladder (Q2368834)

In the paper an edge-reinforced random walk on the ladder \(\mathbb Z\times \{1, 2\}\) is considered. The edges are undirected. They are assigned time-dependent random weights, with all initial edge weights equal to some constant \(a >0.\) In each step, the random walker jumps to a nearest-neighbor vertex with probability proportional to the weight of the traversed edge. Whenever the random walk crosses an edge, its weight is increased by 1. The authors prove that the edge-reinforced random walk on \(\mathbb Z \times \{1, 2\}\) with initial weights \(a>3/4\) is recurrent. The problem is much more subtle than for acyclic graphs. They use the fact that the edge-reinforced random walk on a finite ladder has the same distribution as a random walk in an environment given by random time-independent edge weights. These edge weights are stochastically dependent in a complicated way.