先验判断信息下主-客体极大熵博弈模型  

Game Model of the Maximum Entropy for Subject-object Based on Priori Information

作　　者：陶良彦[1] 刘思峰[1] 方志耕[1] 袁潮清[1] 

机构地区：[1]南京航空航天大学经济与管理学院,江苏南京211106

出　　处：《系统工程》2013年第9期73-78,共6页Systems Engineering

基　　金：国家自然科学基金资助项目(70971064;90924022;71111130211);江苏高校哲学社会科学重点研究基地重大项目(2010JDXM014);南京航空航天大学研究生创新基地(实验室)开放基金资助项目(kfjj120126);中央高校基本科研业务费专项

摘　　要：实践中主体基于对客体的认识进行决策的现象大量存在,可以将此类问题看成主体和客体间的博弈问题,但将客体视作与主体同等思想的传统处理方法并不完善。针对此类问题,构建了主-客体极大熵博弈模型。定义了各局中人的行为准则:主体局中人依据相关经验、统计资料等先验判断信息,采用极大熵准则揣测客体局中人的博弈策略,在此基础上依据自身效用最大化原则选择博弈策略。当期望收益与实际收益差距较大时,利用贝叶斯推理修正客体局中人策略。最后用一个风险控制的案例验证了本文所提方法的有效性和适用性。The phenomenon that the subject makes decision based on the understanding of object is abundant in real life. Such problems can he seen as the game between subject and object, but the traditional approach which thinks the object has the same ideas with the subject is not perfect. In order to solve this problem, the concept of the subject-object player game and the player＇s conduct code are defined. According to relevant priori information, such as experience and statistics, the subject player presumes object player＇s game strategy using maximum entropy criterion. On the basis of the proof for the existence of the object player＇s optimal strategy solution, the subject player＇s game strategy selection method is presented in accordance with the principle of individual interests. When the gap between the expected returns and actual returns is very big, Bayesian correction method is used to amend the object＇s strategy. Finally, the validity and applicability of the proposed method is validated by a risk-controlled case.

关 键 词：博弈 极大熵准则 主-客体局中人 贝叶斯推理 

分 类 号：O225[理学—运筹学与控制论]