An Algebraic Approach to Formulating and Solving Large Models for Sequential Decisions Under Uncertainty
提出一种代数方法，将决策问题用变量和函数紧凑表示，并用简单算法快速求解原本需要数十万端点的决策树模型，适用于管理科学中常用启发式处理的复杂序贯决策问题。
This article presents an algebraic approach to formulating and solving large models for sequential decisions under uncertainty. With this approach, decision analysis optimization methods can be applied to complex decision problems which are generally analyzed in management science practice using heuristics. Using the approach, a decision problem is formulated in terms of decision variables, random variables, and functions relating these variables. This leads to a compact representation, and a simple algorithm can be used to quickly solve algebraic models that would have decision trees with several hundred thousand endpoints. An application to research and development planning illustrates the usefulness of such large sequential decision models.