On Simulating a Markov Chain Stationary Distribution when Transition Probabilities are Unknown
暂无分享,去创建一个
We present an algorithm which, given a n-state Markov chain whose steps can be simulated, outputs a random state whose distribution is within ϵ of the stationary distribution, using O(n)space and O(ϵ-2τ) time, where is a certain “average hitting time” parameter of the chain.
[1] Lyle A. McGeoch,et al. On-Line Algorithms , 1992 .
[2] Peter W. Glynn,et al. Stationarity detection in the initial transient problem , 1992, TOMC.