Markov decision processes in practice
Markov decision processes in practice
复制标题
DOI：
10.1007/978-3-319-47766-4
复制
发表时间：
2017
期刊：
影响因子：
--
通讯作者：
R. Boucherie;N. Dijk
中科院分区：
文献类型：
--
作者：
R. Boucherie;N. Dijk
自从DJ白色开始他的一系列关于马尔可夫决策过程（MDP）的实际应用的调查以来，已经过去了30多年，1，2，3在Martin Puterman关于MDP理论的惊人的书之后的20多年，4和尤金A. Feinberg和Adam Shwartz出版了他们的《马尔可夫决策过程手册：方法与应用》。在过去的几十年里，MDP的实际发展似乎已经停止，因为人们普遍认为MDP在计算上是禁止的。因此，MDP被认为是不现实的，并且超出了许多运筹学从业者的范围。此外，MDP还受到其符号复杂性和概念复杂性的阻碍。因此，MDP通常只在介绍性的运筹学教科书和课程中简要介绍。最近发展的近似技术支持的数值能力大大增加，已解决了部分计算问题，见，例如，第章。本手册的第2和第3条以及其中的参考文献。这本手册表明，出于实际目的而恢复MDP有几个理由：
It is over 30 years ago since DJ White started his series of surveys on practical applications of Markov decision processes (MDP), 1, 2, 3 over 20 years after the phenomenal book by Martin Puterman on the theory of MDP, 4 and over 10 years since Eugene A. Feinberg and Adam Shwartz published their Handbook of Markov Decision Processes: Methods and Applications. 5 In the past decades, the practical development of MDP seemed to have come to a halt with the general perception that MDP is computationally prohibitive. Accordingly, MDP is deemed unrealistic and is out of scope for many operations research practitioners. In addition, MDP is hampered by its notational complications and its conceptual complexity. As a result, MDP is often only briefly covered in introductory operations research textbooks and courses. Recently developed approximation techniques supported by vastly increased numerical power have tackled part of the computational problems; see, eg, Chaps. 2 and 3 of this handbook and the references therein. This handbook shows that a revival of MDP for practical purposes is justified for several reasons: