No label defined (Q3767154)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 4025193
Language Label Description Also known as
default for all languages
No label defined
    English
    No label defined
    scientific article; zbMATH DE number 4025193

      Statements

      0 references
      1987
      0 references
      discounted stochastic game
      0 references
      average reward stochastic game
      0 references
      The book is a monograph devoted to two-person stochastic games. The games are considered mainly under the assumption that the spaces of pure actions of the players and the state space are finite. The monograph acquaints the reader with the history, and with classical and latest results obtained in the theory of stochastic games. These games have the form of an ordered quintuple \(<S\), \(\{A_ s| s\in S\}\), \(\{B_ s| s\in S\},r,p>\), where S and \(A_ s\) and \(B_ s\) for all \(s\in S\) are finite non-empty sets, r is a real function on the set \(H=\{(s,i,j)|\) \(s\in S\), \(i\in A_ s\), \(j\in B_ s\}\), and p is a map p:H\(\to P(S)\) with P(S) the family of probability distributions on the set S. The game can be interpreted in the following way: An initial state \(s_ 1\) known to both players is chosen arbitrarily from the state space S and next, both players are to select, simultaneously and independently of each other, actions \(i_ 1\in A_{s_ 1}\) and \(j_ 1\in B_{s_ 1}\) from their action sets \(A_{s_ 1}\) and \(B_{s_ 1}\) in state \(s_ 1\), respectively. This ends the first stage of the game and then player 2 pays to player 1 the amount \(r_ 1=r(s_ 1,i_ 1,j_ 1)\) described by the payoff function r. Therefore one can see that on the first stage, the players play the zero-sum matrix game in the normal form \(<A_{s_ 1},B_{s_ 1},r(s_ 1,\cdot,\cdot)>\). The situation is repeated on the second and further stages. Namely, a state \(s_ 2\in S\) is chosen according to the transition map p in such a way that the probability of \(s_ 2\) is depended on \(s_ 1\), \(i_ 1\) and \(j_ 1\) from the previous stage, and is equal to \(p(s_ 2| s_ 1,i_ 1,j_ 1)\). Next, after getting the information on \(s_ 2\), the players play the matrix game \(<A_{s_ 2},B_{s_ 2},r(s_ 2,\cdot,\cdot)>\) which gives a payoff \(\tau_ 2=r(s_ 2,i_ 2,j_ 2)\) to player 1, and so on. Let \(r_ 1,r_ 2,r_ 3,..\). be payoffs of player 1 on all the stages. A general payoff for player 1, the sum \[ \sum^{\infty}_{i=1}\beta^{i-1}r_ i\quad (0<\beta <1) \] or one of the limits, \[ \underline{\lim}_{k\to \infty}(\sum^{k}_{i=1}r_ i)/k\quad or\quad \overline{\lim}_{k\to \infty}(\sum^{k}_{i=1}r_ i)/k \] is considered, and player 1 wishes to maximize his general payoff while player 2 wishes to minimize it. This model corresponds to a two- person zero-sum game called discounted stochastic game or average reward stochastic game, respectively. Economic interpretations of those games are obvious. In the book both classes of stochastic games are widely discussed. A reader can become acquainted with problems like different structures of the players' strategies, the existence of a game value and optimal strategies for the players, structural properties of stochastic games, and a review of methods used in the theory. Besides, the book presents algorithms for finding values and optimal strategies in stochastic games, including the latest results in this field. If we add that the recent literature on stochastic games is also shortly discussed, the monograph may be recommended for the readers as deeply introducing into those games.
      0 references

      Identifiers

      0 references
      0 references
      0 references
      0 references
      0 references
      0 references