No label defined (Q4944328)

The authors consider a three-person, nonzero-sum game with a finite number of pure strategies for the three players. Let there be \(r\) pure strategies for player 1, \(s\) pure strategies for player 2 and \(t\) pure strategies for player 3. The payoff to player \(l\) for the pure strategies \(i\) of player 1, \(j\) of player 2 and \(k\) of player 3 is given by \(u^l_{ijk}.\) It is also given that players 2 and 3 adopt the apriori known mixed strategies \(\eta \) and \(\zeta .\) The gain function \( F_1 \) of player 1 with a mixed strategy \(\xi \) is assumed to have the form \[ F_1 (\xi) = \lambda _1\sum _{i=1}^ru_i\xi _i+\lambda _2 \sum_{i=1}^rv_i\xi _iln\frac{1}{\xi _i}, \] where \(u_i =\sum_{j=1}^s\sum_{k=1}^t u^1_{ijk}\eta _j\zeta _k\) and where \(\lambda _1, \lambda _2\) are positive and \(v_i >0 for i =1, 2,\ldots , r.\) The purpose of player 1 is to maximize his gain function by choosing a mixed strategy from the set of mixed strategies satisfying a given linear restriction. The authors present a few results on the optimal and \(\varepsilon \) optimal strategies of player 1 for this problem using the results on entropy optimization derived in the following paper [\textit{S. C. Fang}, and \textit{H. S. J. Tsao}, Z. Oper. Res. 37, 171-186 (1993; Zbl 0777.90026)]. They also present sketches of two algorithms.