SCREMS: Distributed Environments for Stochastic Computation

SCEMS：随机计算的分布式环境

• 批准号: 0422400
• 负责人: Merlise Clyde
• 金额: --
• 依托单位: Duke University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2004
• 资助国家: 美国
• 起止时间: 2004-09-01 至 2007-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0422400&HistoricalAwards=false
• 关键词: SCREMS; Distributed; Environments; Stochastic; Computation

This grant will allow the Institute of Statistics and Decision Sciences(ISDS) at Duke University to purchase equipment for a high-speedcomputing cluster which will be dedicated to support research incomputational statistical sciences. It will be used for five projectsspecifically directed at developing statistical methodology for complex,high-dimensional hierarchical models using distributed computingenvironments. Specific research problems involve sequential importancesampling in latent variable models, combinatorial search for data miningapplications, adaptive stochastic search for high dimensional modelspaces, multivariate nonparametric function estimation, and multivariatespatial-temporal modeling. These problems, which are typical ofapplications in many scientific disciplines, require substantialcomputational power, in both speed and memory, and require development ofnew methodology for efficient implementations in modern distributedcomputing environments. Modern statistical methodology has allowed researchers to propose morecomplex and realistic models for many scientific problems, however,powerful computing environments and sutiable methodology are necessaryto make progress as larger datasets become available. This grant willallow the Institute of Statistics and Decision Sciences (ISDS) at DukeUniversity to purchase a high-speed computing cluster, dedicated tosupport research in computational statistical sciences. Projectsinclude spatio-temporal models to integrate information at multiplelevels, with applications in climate change and ecology, new classes ofmodels for identifying protein biomarkers for early cancer detection,novel algorithms for exploring high-dimensional collections of modelsand combining them, new techniques for modern data mining, and efficientmethodology for sequential analysis of data collected over time. Theseare all areas at the forefront of statistics with broad applicability towidespread problems in the statistics, science and beyond. The proposedcomputing system will be both extensible and upgradeable. Servinginitially the five projects, it will also provide for future expansionas these projects evolve and generate new directions in collaborativeresearch involving the primary investigators. The proposed researchwill have significant immediate impact on education through involvementof Ph.D. students and postdoctoral researchers directly in the proposedresearch, and incorporation of results into graduate courses at DukeUniversity. Results will be disseminated to the statistical and broaderscientific communities via software distributed on the web, direct collaborationwith researchers in other disciplines, conference presentations, andpapers to be published in academic journals.

这笔赠款将允许统计和决策科学研究所（ISDS）在杜克大学购买设备的高速计算集群将致力于支持研究incomputational统计科学。 它将用于五个项目，专门针对开发复杂的，高维层次模型的统计方法，使用分布式计算环境。 具体的研究问题涉及潜在变量模型中的序列重要性，数据挖掘应用的组合搜索，高维模型空间的自适应随机搜索，多变量非参数函数估计和多变量时空建模。 这些问题，这是典型的应用程序在许多科学学科，需要大量的计算能力，在速度和内存，并要求开发新的方法，有效地实现在现代分布式计算环境。 现代统计方法使研究人员能够为许多科学问题提出更复杂和更现实的模型，然而，随着更大的数据集变得可用，强大的计算环境和合适的方法是必要的。这项拨款将允许杜克大学的统计与决策科学研究所（ISDS）购买一个高速计算集群，致力于支持计算统计科学的研究。项目包括在多个层面整合信息的时空模型，以及在气候变化和生态学中的应用，用于识别早期癌症检测的蛋白质生物标志物的新型模型，用于探索高维模型集合并将其组合的新算法，现代数据挖掘的新技术，以及对随时间收集的数据进行顺序分析的有效方法。 这些都是统计学的前沿领域，对统计学、科学和其他领域的普遍问题具有广泛的适用性。 建议的计算系统将是可扩展的和可移植的。 最初服务于五个项目，它还将为未来的扩展提供服务，因为这些项目不断发展，并在涉及主要研究者的合作研究中产生新的方向。 通过博士的参与，这项研究将对教育产生重大的直接影响。 学生和博士后研究人员直接在拟议的研究，并纳入杜克大学的研究生课程的结果。 研究结果将通过网络上的软件、与其他学科研究人员的直接合作、会议报告和在学术期刊上发表的论文传播给统计界和更广泛的科学界。