影响图的扩展综述

A Survey on Extension of Influence Diagrams

影响图是单决策人基于不确定信息表示和求解复杂决策问题的图模型,已成为一种流行的标准建模工具。由于决策环境纷繁复杂,决策问题多种多样,没有任何一种建模工具能够普遍适用于各种决策问题。为了表示复杂决策问题,增强影响图的表达能力,研究者对影响图进行了多种扩展。从建模无限制决策问题、非对称决策问题、涉及连续变量的决策问题、涉及非精确变量的决策问题及多Agent决策问题等方面对影响图的扩展进行了较为全面的回顾与分析,指出了有关影响图的进一步研究方向。关键词:决策;影响图;无限制决策;非对称决策;

Influence diagrams （IDs） are graphical models for representing and solving complex decision-making problems based on uncertain information for a single decision maker. IDs have become a popular and standard modeling tool. However, there are various decision-making problems in real-world, so the need to represent a vari- ety of complex problems leads to the extensions of influence diagram methodology to increase their representation power. This paper reviews the extensions associated with influence diagrams for representing unconstrained, asym- metric, continuous, qualitative and multi-agent decision problems, and describes the relevant further researches on IDs.