Acquisition of Linux Cluster to Meet Modern Computational Needs for Statistical Research at University of Georgia

格鲁吉亚大学统计系的教师和学生参与了令人兴奋的研究项目，其中许多是计算密集型的。 目前的计算设备不足以满足日益增长的研究需求，包括功能性神经成像分析，基因组生态学建模，贝叶斯经验似然方法，符号数据方法和高维数据简化。 理论研究和跨学科应用都涉及计算工作。 对基因组和大脑成像的研究需要大的数据存储容量以及计算速度。 研究新的统计方法往往需要大量的、计算密集型的模拟。例如，贝叶斯经验似然应用的后验分布的计算是相当计算密集的，并且很适合于并行计算环境。 高维数据缩减的一种开发方法涉及序列二次规划，当并行编程时，它可以快很多倍。 所列举的每一个研究项目都将从新的项目组中受益匪浅。 许多博士生都参与了这些项目;在最现代化的计算设施中进行培训将提高学生毕业后的就业前景，更好的计算资源将使该部门在招聘和留住优质学生方面更具吸引力。 格鲁吉亚大学统计系的研究人员需要购买最先进的Linux集群，以充分发挥其工作潜力。 计算密集型的研究项目，为六名教员列举在项目说明。 这些项目需要理论和跨学科的研究，包括开发用于脑成像的新方法，基因组生态学建模以及气候时间序列的推断。 理论的进步和现实世界的应用形成了一种共生关系，因为每一种活动都是相互补充的。 对新统计方法的研究通常涉及计算密集型模拟和/或非常大的数据集。 例如，为了更好地理解人类大脑的工作，功能性神经成像的研究涉及包含许多个体在各种刺激下的图像的数据集。 方法论必须对个体之间、刺激之间以及每个大脑内部的差异进行建模，以确定功能和活动之间的关系。 同样，对人类基因组的研究涉及大量数据和计算密集型分析。 密集的计算机工作可以为理论研究提供新的方向。 该部门目前的计算机设备限制了目前手头项目的范围，不足以满足年轻而活跃的教师所设想的新方向。

The faculty and students in the University of Georgia Statistics Department are involved in exciting research projects, many of which are computationally intensive. The current computing equipment is inadequate to meet the growing needs of the research, which includes functional neuro-imaging analyses, modeling the ecology of the genome, Bayesian empirical likelihood methods, methodology for symbolic data, and high-dimensional data reduction. Both theoretical research and interdisciplinary applications involve computational work. Research into the genome and brain imaging requires large data storage capacities as well as computational speed. Research into new statistical methodologies often requires extensive, computationally intensive simulations. For example, the calculation of posterior distributions for Bayesian Empirical Likelihood applications is quite computationally intensive, and lends itself well to a parallel computing environment. One of the developing methods for high-dimensional data reduction involves sequential quadratic programming, which could be many times faster when programmed in parallel. Each of the enumerated research projects will benefit greatly from the new cluster. Many PhD students are involved in the projects; training in the most modern computing facilities will enhance students' job prospects after graduation, and better computing resources will make the department more attractive for recruiting and retaining top-quality students. The acquisition of a state-of-the-art Linux cluster is necessary for the researchers at the University of Georgia Statistics Department to realize the full potential of their efforts. Computationally intensive research projects for six faculty members are enumerated in the project description. These projects entail both theoretical and interdisciplinary research, including development of new methodologies for use in brain imaging, modeling of the ecology of the genome, and inference about climatological time series. Theoretical advances and real-world applications form a symbiotic relationship, as each activity nourishes the other. Often research into new statistical methodologies involves computationally intensive simulations and/or very large datasets. For example, to better understand the workings of the human brain, research in functional neuro-imaging involves datasets containing images for many individuals under various stimuli. Methodologies must model variation across individuals and across stimuli, as well as within each brain, to determine relationships between function and activity. Similarly, research into the human genome involves enormous amounts of data and computationally intensive analyses. New directions in theoretical research can be suggested by intensive computer work. The current computing equipment in the department is limiting the scope of the projects currently in hand, and is inadequate for the new directions envisioned by a young and active faculty.