Special Meeting: Workshop on Algorithms for Modern, Massive Datasets

特别会议：现代海量数据集算法研讨会

• 批准号: 0532668
• 负责人: Gene Golub
• 金额: $ 1.5万
• 依托单位: Stanford University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2005
• 资助国家: 美国
• 起止时间: 2005-08-15 至 2006-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0532668&HistoricalAwards=false
• 关键词: Special; Meeting; Workshop; Algorithms; Modern

This special meeting will bring together people from the ComputationalStatistics and Numerical Linear Algebra communities with people from theTheoretical Computer Science community to promote cross-fertilization ofideas. This will hopefully lead to a better understanding of modernmassive data sets as well as invention of new techniques and algorithmsfor their modeling and analysis. Most conferences and workshops oninformation retrieval and data mining are dominated by studies oftraditional data sets and traditional algorithmic techniques.Nevertheless, modern data sets have become extremely complex and haverendered traditional approaches for processing them ineffective in manyinstances. The difficulties occur primarily in two ways: modern data setsare often high-dimensional and massive, and thus novel dimensionalityreduction or sampling-based techniques have to be employed in processingthem, and modern data sets often have complicated nonlinear structures,and must be modeled with appropriate mathematical objects such as tensors,symmetric spaces, Lie groups, etc. This special meeting will study datasets with such complex structures and explore novel techniques foranalyzing such data.The techniques that will be discussed at the meeting have applicability innumerous fields of practical and important interest. For example:(1) Applications in Bioinfomatics and the Medical Sciences includeimproved methods to analyze large microarray and high-throughput chemicaldata sets for developing new medications, identifying novel gene products,elucidating protein folding pathways, and the improved detection andclassification of cancer.(2) Applications in the Social Sciences include improved methods to modelthe combinatorial structure of large social networks; this has been usedand will continue to be used to, e.g., identify potential terrorist cellshidden in a communications network.(3) Applications in the Geosciences include remote sensors in air-andspacecraft that produce more data than can currently be stored, but whichmust be analyzed for environmental planning, weather forecasting, andpublic health contamination issues.(4) Applications in Computer Vision include novel ways to identify noisyimages of targets and faces for improved target identification andindividual identification in more realistic settings.

这次特别会议将把计算统计和数值线性代数社区的人与理论计算机科学社区的人聚集在一起，以促进思想的交叉施肥。 这将有望导致更好地理解现代大规模数据集以及发明新的技术和算法来建模和分析。 大多数关于信息检索和数据挖掘的会议和研讨会都以传统数据集和传统算法技术的研究为主。然而，现代数据集已经变得极其复杂，并且在许多情况下使处理它们的传统方法变得无效。 困难主要表现在两个方面：现代数据集通常是高维和海量的，因此必须采用新的降维或基于采样的技术来处理它们，并且现代数据集通常具有复杂的非线性结构，并且必须用适当的数学对象（例如张量、对称空间、李群，这次特别会议将研究具有如此复杂结构的数据集，并探索分析此类数据的新技术。会议将讨论的技术在许多实际和重要的领域具有适用性。例如：（1）在生物信息学和医学科学中的应用包括改进的方法来分析大型微阵列和高通量化学数据集，用于开发新药，鉴定新的基因产物，阐明蛋白质折叠途径，以及改进癌症的检测和分类。(2)在社会科学中的应用包括改进的方法来模拟大型社交网络的组合结构;这已经被使用，并将继续被用于，例如，在通讯网络中找出潜在的恐怖分子(3)在地球科学中的应用包括空气中的遥感器和航天器，它们产生的数据比目前存储的数据要多，但必须对环境规划、天气预报和公共卫生污染问题进行分析。(4)在计算机视觉中的应用包括识别目标和人脸的噪声图像的新方法，以改进目标识别和更真实环境中的个人识别。