Linear programming in a Markov chain: notes on linear programming and extensions-part 59.
George B. Dantzig, Philip Wolfe · 1962
Abstract : An infinite Markov process with a finite number of states is considered in which the transition probabilities for each stage range independently over sets that either are finite or are convex polyhedra. A finite computational procedure is given for choosing those transition probabilities which minimize appropriate functions of the resulting equilibrium probabilities.