摘要
期望均衡是博弈局中人或局外人对于博弈均衡点的一致期望,强调互利共赢,它要求博弈群体的每个成员对期望均衡点有一个共同的预期.显然,基于纳什均衡的帕累托优化组合策略是一个比纳什均衡更有效的期望均衡.要实现期望均衡,可采用局中人参与的训练与学习使得群体的预期一致,也可采用第三方过滤器来达到期望目标.在期望均衡的概率分布下,个体行为的偏离不能比均衡态取得更多收益.否则,训练就是无效的,第三方过滤器就是不公平的.
Expecting equilibrium is an identical expecting on game equilibrium point from the players in the game or the third parties out of the game. It pays attention to mutual success, and demands all of the players to have the same expecting. Obviously, a Pareto dominating strategy profile based on Nash equilibrium is an expecting equilibrium which is more efficient than Nash equilibrium. To achieve an expecting equilibrium, we can use training and learning to make the players have the same expecting, while we can use the third-party filter to realize the expecting goal. With the probability distribution of expecting equilibrium, a departure of personal behavior from the equilibrium can not gain more utility. Otherwise, training will be of no effect and the third-party filter will be unfair.
出处
《数学的实践与认识》
CSCD
北大核心
2010年第18期36-42,共7页
Mathematics in Practice and Theory
关键词
博弈学习
期望均衡
训练
第三方过滤器
game learning
expecting equilibrium
training
third-party filter