Propp-Wilson algorithms and finitary codings for high noise Markov random fields (Q2709847)

It is known, that a random field with finite state space \(S\) indexed by the integer lattice \(Z^{d}\) is a random mapping \(X:Z^{d}\to S,\) or equivalently is a random element of \(S^{Z^{d}}.\) The present paper is focused on so-called Markov random fields, characterized by having a dependency structure which only propagates via interactions between nearest neighbours in \(Z^{d}.\) It is specialized further to Markov random fields satisfying a certain high noise assumption. The results presented for such random fields are twofold. First, it is devised a so-called Propp-Wilson algorithm for computer simulation of the random field. Second, it is used the existence and some properties of such an algorithm to prove that the high noise assumption implies a certain rather strong ergodic property, known as finitary codability, of the random field. The authors combine two previous works in this paper, the first being by the first author and \textit{K. Nelander} [Scand. J. Stat. 26, No. 3, 395-411 (1999; Zbl 0944.60059)], and the second by \textit{J. van den Berg} and the second author [Ann. Probab. 27, No. 3, 1501-1522 (1999; Zbl 0968.60091)], to obtain the above mentioned results.