Mixing times for neighbour transposition shuffles on graphs
Mixing times for neighbour transposition shuffles on graphs
复制标题
图上相邻转置洗牌的混合时间
DOI:
--
复制
发表时间:
2011
期刊:
影响因子:
--
通讯作者:
Stefan Erikshed
中科院分区:
文献类型:
--
作者:
Stefan Erikshed
This thesis will treat Markov chains on the symmetric group Sn, i.e. the set of permutations of n distinct objects. Markov chains with the state space Sn are often referred to as card shuffling chains. The state space is thus the reorderings of a deck of n cards. A certain type of card shuffling chains is considered here; neighbour transpositions on graphs. This is a generalization of ordinary random transpositions shuffle. Each step of the ordinary random transpositions shuffle consists of randomly selecting any pair of cards in the deck and then switch their places. Neighbour transpositions on a graph means that the n cards is placed on the vertices of a n-vertex graph. At each step a neighbour pair of cards in the graph (i.e. two cards at positions connected with an edge) is selected and transposed. If the graph is connected, the deck will eventually be well mixed. In other words, the distribution of the chain converges to uniformity on Sn. This thesis deals with the rate of convergence to uniformity for card shuffling on two families of graphs, lollipop graphs and random graphs, G(n, p). More precisely, bounds on the mixing time of these shuffles is determined. The mixing time is the number of steps of the Markov chain until it is close to its stationary uniform distribution. As usual when dealing with convergence rate problems we let |Sn| → ∞, yielding asymptotic results in n. Lower and upper bounds, both of order n log n, on the mixing time for neighbour transpositions on lollipop graphs is derived. Further, lower bounds of order n log n, on the mixing time for neighbour transpositions on connected random graphs is established. Upper bounds of the same order is proved for random graphs with bounded diameter.