报告时间：2021年12月28日

               上午10:00-11:00

报告地点：腾讯会议

              （会议ID：642 562 472）

报告嘉宾：晏挺

报告主题：Regularized maximum likelihood in the beta model for large and sparse networks

报告摘要

Regularized maximum likelihood in the beta model for large and sparse networks

The beta model is a powerful tool for modeling network generation driven by node degree heterogeneity.  It is simple yet expressive nature particularly well-suits large and sparse networks, where even moderately complex models might be infeasible to fit due to very few nonzero observations and computational challenge. However, simple as this model is, our theoretical understanding remains rather limited.  Also, available computation method for fitting this model remains unscalable.  In a big-data era, substantial improvements are urgently needed for the beta model. Our paper brings several major refinements and improvements to the methodology and theory of the beta model: 1. we propose a new L2 penalized MLE scheme; we design a novel algorithm that can comfortably handle sparse networks of millions of nodes, sharply contrasting the best existing tools that could only deal with thousands of nodes; 2. we present much stronger error bounds on beta-models under much weaker assumptions than existing literature; we also prove the first resolution-limit bound and new normality results; 3. we apply our method to analyze a huge COVID-19 knowledge graph and discover very meaningful results.

个人简介

晏挺，现任华中师范大学数学与统计学学院教授，目前的主要研究方向有网络数据分析，成对比较数据分析等，主持了多项国家自然科学基金项目。曾在乔治华盛顿大学从事博士后研究，于2013年进入华中师范大学工作，入选了湖北省楚天学者计划。在Annals of Statistics, Journal of the American Statistical Association, Biometrika等统计学期刊上发表了三十余篇论文。