Advances in approximate and adaptive dynamic programming

近似和自适应动态规划的进展

• 批准号: 5527-2007
• 负责人: Puterman, Martin
• 金额: $ 3.67万
• 依托单位: University of British Columbia
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2008
• 资助国家: 加拿大
• 起止时间: 2008-01-01 至 2009-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=389156
• 关键词: Advances; approximate; adaptive; dynamic; programming

Research carried out under this proposal will advance theory and application.  It will focus on complex problems in sequential decision making under uncertainty paying special attention to advancing methods for solving large problems (approximate dynamic programming) and methods for learning system parameters in the course of the decision making process (adaptive dynamic programming).  For approximate dynamic programming, it will provide new bounds on approximate solutions along with several theoretical and practical results.  Novel applications of approximate dynamic programming to health care management and online auctions will complement the theoretical advances and provide further insight into the practical aspects of using these methods.  Adaptive dynamic programming methods will be applied to problems in prostate cancer treatment and selling through Internet auctions.   The applications of this research to healthcare issues address important management challenges concerning the efficient use of scarce resources. Innovations in this area will significantly impact the number one issue facing Canadians today; timely access to quality medical care.

根据这一建议进行的研究将促进理论和应用。重点研究不确定条件下序列决策中的复杂问题，重点提出求解大问题的方法（近似动态规划）和决策过程中系统参数的学习方法（自适应动态规划）。对于近似动态规划，它将提供近似解的新边界以及一些理论和实践结果。近似动态规划的医疗保健管理和在线拍卖的新应用将补充理论的进步，并提供进一步深入了解使用这些方法的实际方面。自适应动态规划方法将应用于前列腺癌治疗和通过互联网拍卖销售的问题。本研究在医疗保健问题上的应用解决了有关有效利用稀缺资源的重要管理挑战。这一领域的创新将对加拿大人今天面临的头号问题产生重大影响；及时获得高质量的医疗服务。