Created by Knut M. Synstadfrom the Noun Project

Dobrushin's Ergodic Coefficient

Definition (Dobrushin Coefficient)

Consider a countable state space Ω\Omega and MC {Xi}i=1\{ X_{i} \}_{i=1}^{\infty} with one-step transition matrix PP. The Dobrushin Ergodic Coefficient is δ(P)=mini,kjXmin(Pij,Pkj)\delta(P)=\min_{i,k}\sum_{j\in\mathbb{X}}\min(P_{ij},P_{kj})

Theorem (Dobrushin Convergence Theorem)

For any two probability measures, π,π\pi,\pi' it follows that πPπP(1δ(P))ππ\|\pi P-\pi'P\|\le(1-\delta(P))\|\pi-\pi'\|hence, πk+1=πkP(=:T(πk))\pi_{k+1}=\pi_{k}P\quad(=:T(\pi_{k}))converges to the unique invariant probability measure π\pi^{*} πnπ2(1δ(P))n\|\pi_{n}-\pi^*\|\le2(1-\delta(P))^{n}