III: Small: High-Performance Complex Processing of Continuous Uncertain Data

本项目的目标是设计和开发一个数据管理系统，通过返回查询输出的全概率分布来支持对连续不确定数据的查询处理，并对这种处理进行性能优化。该项目包括四个方面：(1)使用传统关系模型和数组模型支持连续的不确定数据处理；(2)使用新的统计图形模型处理连续不确定数据处理中出现的复杂相关性；(3)通过探索诸如高斯过程和函数内插等高级技术，除标准查询操作外，支持任意用户定义的函数；以及(4)开发一个原型系统，并使用现实世界的应用程序对其进行评估。预期结果包括统计模型和技术、数据存储方案、查询处理和优化技术，以及完全支持对连续不确定数据的查询处理的公开可用的原型。该项目的结果可以使恶劣天气监测和计算天体物理学等应用程序以及更广泛的科学界受益。由于龙卷风探测等应用可能会根据衍生信息触发行动，因此表征产出不确定性的能力可能会产生重大的社会影响。该项目还将研究和教育与课程开发相结合，并通过大学外展和CRA的分布式导师计划吸引妇女参与研究。该项目的结果在该项目的网站上公布：http://claro.cs.umass.edu.

The objective of this project is to design and develop a data management system that supports query processing on continuous uncertain data by returning a full probability distribution of query output and optimizes such processing for performance. This project includes four thrusts: (1) supporting continuous uncertain data processing using both the traditional relational model and the array model; (2) addressing complex correlation that arises in continuous uncertain data processing using new statistical graphical models; (3) supporting arbitrary user-defined functions, besides standard query operations, by exploring advanced techniques such as Gaussian processes and functional interpolation; and (4) developing a prototype system and evaluating it using real-world applications. Expected results include statistical models and techniques, data storage schemes, query processing and optimization techniques, and a publicly available prototype to fully support query processing on continuous uncertain data. The results of the project can benefit applications such as severe weather monitoring and computational astrophysics, as well as the broader scientific community. Since applications such as tornado detection may trigger actions based on derived information, the ability to characterize uncertainty of output may result in significant social impacts. This project also integrates research and education with curriculum development and engaging women in research through college outreach and CRA's distributed mentor program. The results of the project are disseminated at the project web site: http://claro.cs.umass.edu.