An Approach for Reducing the Graphical Model and Genetic Algorithm for Computing Approximate Nash Equilibrium in Static Games
An Approach for Reducing the Graphical Model and Genetic Algorithm for Computing Approximate Nash Equilibrium in Static Games
复制标题
DOI：
10.1007/s10846-010-9419-6
复制
发表时间：
2010-11
影响因子：
3.3
通讯作者：
Weiyi Liu;Kun Yue;Jin Li;Ning Song;Li Ding
中科院分区：
文献类型：
--
作者：
Weiyi Liu;Kun Yue;Jin Li;Ning Song;Li Ding
本文研究了静态多智能体博弈中图模型的简化方法和求解近似纳什均衡的遗传算法。为了描述各Agent策略之间的关系，提出了影响度和策略依赖度的概念。基于这些概念，给出了一种简化图模型的方法。对于离散化的混合策略，给出了离散度与近似度之间的关系。基于后悔度，给出了求解近似纳什均衡的遗传算法。实验结果表明，该算法具有效率高、平衡误差小的特点.
In this paper, an approach for reducing the graphical model and a genetic algorithm for computing the approximate Nash equilibrium in a static multi-agent game is studied. In order to describe the relationship between strategies of various agents, the concepts of the influence degree and the strategy dependency are presented. Based on these concepts, an approach for reducing a graphical model is given. For discretized mixed strategies, the relationship between the discrete degree and the approximate degree is developed. Based on the regret degree, a genetic algorithm for computing the approximate Nash equilibrium is given. Experimental results indicate the genetic algorithm has high efficiency and few equilibrium errors.