ALGORITHMS: Distributed Multivariate Integration in a Problem Solving Environment

算法：问题解决环境中的分布式多元集成

• 批准号: 0203776
• 负责人: Elise deDoncker
• 金额: $ 34.85万
• 依托单位: Western Michigan University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2002
• 资助国家: 美国
• 起止时间: 2002-05-15 至 2005-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0203776&HistoricalAwards=false
• 关键词: ALGORITHMS; Distributed; Multivariate; Integration; Problem

This proposal seeks to extend the PARINT software package for parallel integration, furnishing it with the characteristics of a Problem Solving Environment. The main goals of the project are to offer a complete set of tools for composing problem secifications, compiling and running applications, while providing state-of-the-art problem solving power to the end user. To this end, we propose adding new techniques for extending the PARINT problem domain, e.g., to allow for a variety of integration regions, high dimensions, and singular integrated behavior. These include a transformation interface, new distributed Quasi-Monte Carlo strategies and extapolation methods. Specific contributions are in the development of numerical techniques and evaluation of parallel integration algorithms. Important issues in parallel and distributed computations such as load balancing are addressed, with an emphasis on loosely coupled and heterogeneous systems. Other contributions are in the areas of user interfaces, visualization and web computing.

该提案旨在扩展 PARINT 软件包以实现并行集成，为其提供问题解决环境的特征。该项目的主要目标是提供一整套用于编写问题细分、编译和运行应用程序的工具，同时为最终用户提供最先进的问题解决能力。为此，我们建议添加新技术来扩展 PARINT 问题域，例如，允许各种集成区域、高维度和奇异集成行为。其中包括转换接口、新的分布式准蒙特卡罗策略和外推方法。 具体贡献在于数值技术的开发和并行积分算法的评估。解决了并行和分布式计算中的重要问题（例如负载平衡），重点是松散耦合和异构系统。其他贡献涉及用户界面、可视化和网络计算领域。