Sampling for Statistical Inference on Network Data

网络数据统计推断的采样

• 批准号: 1106796
• 负责人: Yuguo Chen
• 金额: $ 18.01万
• 依托单位: University of Illinois at Urbana-Champaign
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-07-01 至 2015-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1106796&HistoricalAwards=false
• 关键词: Sampling; Statistical; Inference; Network; Data

Network structures arise in modeling a wide variety of systems in sciences and engineering. One of the most important classes of network models is random graphs. The goal of this project is to study two challenging problems in network analysis: one is on sampling random graphs with a given degree sequence, and the other is on detecting structures in networks. For the first problem, a new sequential sampling method is proposed which samples the networks from an approximate uniform distribution. These samples can be used to estimate the distribution of any test statistic and count the number of networks with given vertex degrees. For the second problem, new Monte Carlo algorithms are proposed to detect network structures, including highly connected subgraphs and network motifs.The investigator develops innovative Monte Carlo techniques for simulating random graphs and detecting network structures. These methods can be used to analyze social interaction patterns, identify network motifs, extract densely connected molecular modules, and much more. Identifying network structures are scientifically important because they may correspond to a group of proteins that interact with each other at the same time or a functional unit that plays a key role in biological regulation networks. The research provides an ideal opportunity for involvement of students with a broad range of background and interests. The algorithms developed from this research will be incorporated into relevant courses.

网络结构出现在科学和工程中的各种系统建模中。网络模型中最重要的一类是随机图。本项目的目标是研究网络分析中两个具有挑战性的问题：一个是具有给定度序列的随机图的抽样，另一个是网络结构的检测。对于第一个问题，提出了一种新的序贯抽样方法，该方法从近似均匀分布中对网络进行抽样。这些样本可用于估计任何检验统计量的分布，并计算具有给定顶点度的网络的数量。对于第二个问题，提出了新的Monte Carlo算法来检测网络结构，包括高度连通的子图和网络motifs.The研究者开发了创新的Monte Carlo技术来模拟随机图和检测网络结构。这些方法可用于分析社会互动模式，识别网络基序，提取密集连接的分子模块等等。识别网络结构在科学上很重要，因为它们可能对应于一组同时相互作用的蛋白质，或者在生物调节网络中起关键作用的功能单元。这项研究为具有广泛背景和兴趣的学生参与提供了理想的机会。从这项研究中开发的算法将被纳入相关课程。