TOPICS
Search

Markov Chain Cutoff


Markov chain cutoff is an abrupt asymptotic change of the distance to stationarity from nearly maximal to nearly zero. For a sequence of finite chains, let d_n(t)=max_(x)d_(TV)(P_n^t(x,·),pi_n). Cutoff at times t_n means that, for every fixed 0<epsilon<1, d_n((1-epsilon)t_n)->1 and d_n((1+epsilon)t_n)->0, with integer times rounded when necessary. The same definition uses the transition semigroup in continuous time. It is stronger than knowing only the order of the mixing time.


See also

Glauber Dynamics, Markov Chain, Mixing Time, Total Variation Distance

Explore with Wolfram|Alpha

References

Levin, D. A. and Peres, Y. Markov Chains and Mixing Times, 2nd ed. With contributions by E. L. Wilmer. Providence, RI: American Mathematical Society, 2017. https://doi.org/10.1090/mbk/107.

Cite this as:

Weisstein, Eric W. "Markov Chain Cutoff." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/MarkovChainCutoff.html

Subject classifications