Game-Theoretic Studies on Information Networks and Distributed Systems

信息网络和分布式系统由许多独立的用户和组织共享，这些用户和组织可能是独立/非合作的决策者。可以预期，如果每个决策者通过非合作决策单方面追求自己的绩效目标，则网络的整体性能将被引导到全面改善。此外，预计相互独立的决策，即非合作优化，可能会带来“公平”的情况，使所有决策者获得的利益可能比其他一些决策方式更差，如囚徒困境和布雷斯悖论。我们把这种公用事业的降级称为“悖论”。也就是说，非合作决策可能是帕累托无效率的。在本研究中，我们寻找可能是帕累托有效的，但可能反映非合作决策公平性的方案。

Information networks and distributed systems are shared by a number of independent users and organizations who may be independent/noncooperative decision makers. It may be expected that the entire performance of the networks will be guided to overall improvement, if each decision maker pursues its own performance objective unilaterally by means of the noncooperative decisions. In addition, it is anticipated that mutual independent decision making, i.e. noncooperative optimization may bring about the `fair' situations such that all decision makers may receive worse benefits than some other ways of decision making may do, as exemplified by the prisoners' dilemma and the Braess paradox. We call them such degradations of utilities 'paradoxes.' That is, the noncoopwerative decision making may be Pareto inefficient. In this study, we have searched the scheme that may be Pareto efficient but may reflect the fairness of the noncoopertive decision making.