Coding Markov chains from the past (Q1196317)

Let \((X,T,{\mathcal A},\mu)\) be a dynamical system and \(P\) be a generating partition. Viewing the components of \(P\) as corresponding to symbols, \(X\) may be regarded as the set of doubly-infinite sequences of elements of the finite alphabet \(P\), with \(T\) the shift map. This gives a symbolic stochastic process \((X,T,{\mathcal A},\mu,P)\), or just \((P,\mu)\). A factor map \(\varphi\) between two such processes \((P,m_ P)\) and \((Q,m_ Q)\) is said to be \textit{unilateral} if \(\varphi^{-1}(Q)\) is \(\bigvee^ \infty_{n=0}T^{-n}\) \(P\)-measurable. The author defines the notion of ``unilaterally finitely determined'' (UFD) process (containing all mixing Markov chains with finitely many states) and shows that every UFD process is a unilateral factor of every ergodic process of equal or greater entropy. This extends the work of Sinai for independent processes and that of Ornstein and Weiss for mixing Markov chains in which all transition probabilities are positive. The proofs use Rudolph's theory of joinings as developed by Rothstein and Burton.