Approximating the stationary probability of a state in a Markov chain through Markov chain Monte Carlo techniques is, in general, inefficient. Standard random walk approaches require O(t/p(v)) operations to approximate the probability p(v) of a state v in a chain with mixing time t, and even the best available techniques still have complexity O(t1.5/p(v)0.5); and since these complexities depend inversely on p(v), they can grow beyond any bound in the size of the chain or in its mixing time. In this paper we show that, for time-reversible Markov chains, there exists a simple randomized approximation algorithm that breaks this “small-p(v) barrier”.
On approximating the stationary distribution of time-reversible Markov chains / M. Bressan, E. Peserico, L. Pretto (LEIBNIZ INTERNATIONAL PROCEEDINGS IN INFORMATICS). - In: Symposium on Theoretical Aspects of Computer Science / [a cura di] R. Niedermeier, B. Vallée. - [s.l] : Schloss Dagstuhl- Leibniz-Zentrum fur Informatik GmbH, Dagstuhl Publishing, 2018. - pp. 18:1-18:14 (( Intervento presentato al 35. convegno STACS tenutosi a Caen nel 2018 [10.4230/LIPIcs.STACS.2018.18].
On approximating the stationary distribution of time-reversible Markov chains
M. BressanPrimo
;
2018
Abstract
Approximating the stationary probability of a state in a Markov chain through Markov chain Monte Carlo techniques is, in general, inefficient. Standard random walk approaches require O(t/p(v)) operations to approximate the probability p(v) of a state v in a chain with mixing time t, and even the best available techniques still have complexity O(t1.5/p(v)0.5); and since these complexities depend inversely on p(v), they can grow beyond any bound in the size of the chain or in its mixing time. In this paper we show that, for time-reversible Markov chains, there exists a simple randomized approximation algorithm that breaks this “small-p(v) barrier”.File | Dimensione | Formato | |
---|---|---|---|
LIPIcs-STACS-2018-18.pdf
accesso aperto
Tipologia:
Publisher's version/PDF
Dimensione
606.17 kB
Formato
Adobe PDF
|
606.17 kB | Adobe PDF | Visualizza/Apri |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.