Dependent percolation and colliding random walks
โ Scribed by Peter Winkler
- Publisher
- John Wiley and Sons
- Year
- 2000
- Tongue
- English
- Weight
- 315 KB
- Volume
- 16
- Category
- Article
- ISSN
- 1042-9832
No coin nor oath required. For personal study only.
โฆ Synopsis
Let G be a connected, undirected graph and
Let N N be the nonnegative quadrant of the plane grid, and H the subgraph of N N induced by the sites i j for which X i = Y j . We say that G is "navigable" if with probability greater than 0, the origin belongs to an infinite component of H. We determine which finite graphs are navigable, in particular that K 4 , the complete graph on four nodes, is navigable but K 3 is not. Navigability of G is equivalent to the statement that with positive probability, two tokens taking random walks on G can be moved forward and backward along their paths, and ultimately advanced arbitrarily far, without colliding. The problem is generalized to finite-state Markov chains, and a complete characterization of navigable chains is given. Similar results have been obtained simultaneously and independently by Balister, Bollobรกs and Stacey, using different methods; our classification theorem relies on a surprising diamond lemma which may be of independent interest.
๐ SIMILAR VOLUMES
Consider a graph G and a random walk on it. We want to stop the random walk at certain times (using an optimal stopping rule) to obtain independent samples from a given distribution ฯ on the nodes. For an undirected graph, the expected time between consecutive samples is maximized by a distribution
The N-cube is a graph with 2 N vertices and N 2 Ny1 edges. Suppose indepen- dent uniform random edge weights are assigned and let T be the spanning tree of minimal ลฝ . y 1 N ฯฑ y3 total weight. Then the weight of T is asymptotic to N 2 ร i as N ยช ฯฑ. Asymp-is1 totics are also given for the local stru