Collaborative Research: Novel and Unified Statistical Learning Procedures for Massive Dynamic Multiple-Input, Multiple-Output Networks

协作研究：大规模动态多输入多输出网络的新颖且统一的统计学习程序

• 批准号: 1521761
• 负责人: Chunming Zhang
• 金额: $ 4.39万
• 依托单位: University of Wisconsin-Madison
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-09-01 至 2017-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1521761&HistoricalAwards=false
• 关键词: Collaborative; Research; Novel; Unified; Statistical

This project is to develop an innovative, unified, and flexible methodology to model the input-output transformations in massive multiple-input, multiple-output (MIMO) networks. With the ability of modern technology to gather large volume of high-dimensional and inhomogeneous data, traditional techniques have become inadequate or incapable of modeling the underlying MIMO systems from the data. Novel statistical approaches and methods are needed for extracting knowledge from complex, large, inhomogeneous network data sets. This project seeks to develop new modeling techniques, statistical learning tools, and computational approaches geared towards gaining knowledge arising from different aspects of network data. This development starts with a novel functional dynamic model that accounts for inherent characteristics of network data, such as node latencies, dynamics associated with network coupling, etc. Next, the investigators will develop statistical inference tools to test and identify underlying inherent network connectivity. Furthermore, the investigators will develop statistical learning techniques for network state identification and change-point detection. This project will significantly enrich the toolkit for the analysis of massive network data by providing an alternative data-driven approach. These new tools are expected to provide new solutions to the existing problems and novel and creative solutions to unsolved problems. The research topics address many fundamental problems, for example dynamic modeling, parsimonious network representation, and network learning, for complex MIMO networks when the dimensionality (e.g., the number of nodes) and the sample size grow. It will enhance understanding of existing techniques in the big data context. More importantly, advanced modeling methodology and data analytical tools will be developed with fast-to-implement algorithms, which allow researchers and practitioners to explore, understand, and eventually reverse-engineer complex networks. The investigators will interact with researchers from different fields to derive and validate hypotheses and refine the design of experiments to provide deeper insight into system internals for potential scientific discovery.

该项目旨在开发一种创新、统一且灵活的方法来对大规模多输入多输出（MIMO）网络中的输入输出转换进行建模。随着现代技术能够收集大量高维和非均匀数据，传统技术已经变得不足或无法根据数据对底层 MIMO 系统进行建模。需要新的统计方法和方法来从复杂、大型、非均匀的网络数据集中提取知识。该项目旨在开发新的建模技术、统计学习工具和计算方法，旨在获取网络数据不同方面的知识。这一开发始于一种新颖的功能动态模型，该模型考虑了网络数据的固有特征，例如节点延迟、与网络耦合相关的动态等。接下来，研究人员将开发统计推理工具来测试和识别潜在的固有网络连接性。此外，研究人员将开发用于网络状态识别和变化点检测的统计学习技术。 该项目将通过提供另一种数据驱动的方法来显着丰富用于分析海量网络数据的工具包。这些新工具有望为现有问题提供新的解决方案，并为未解决的问题提供新颖且创造性的解决方案。研究主题解决了复杂 MIMO 网络在维度（例如节点数量）和样本量增长时的许多基本问题，例如动态建模、简约网络表示和网络学习。它将增强对大数据背景下现有技术的理解。更重要的是，先进的建模方法和数据分析工具将通过快速实施的算法来开发，使研究人员和从业者能够探索、理解并最终对复杂网络进行逆向工程。研究人员将与来自不同领域的研究人员互动，推导和验证假设，并完善实验设计，为潜在的科学发现提供对系统内部结构的更深入的了解。